The Neo4j Developer Manual v3 - we-yun.com 3.0 Docs/neo4j-developer... · License: Creative Commons...

The Neo4j Developer Manualv3.0

Table of ContentsIntroduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1. Neo4j highlights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2. Graph Database Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Get started. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3. Install Neo4j . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

4. Get started with Cypher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Cypher Query Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

5. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

6. Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

7. General Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

8. Reading Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

9. Writing Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

10. Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

11. Schema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

12. Query Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

13. Execution Plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

14. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

15. Getting Started. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

16. Driver. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

17. Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221

18. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

19. Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

20. Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

HTTP API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

21. Transactional Cypher HTTP endpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

22. Authentication and Authorization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238

Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241

23. Calling procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

24. Built-in procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244

25. User-defined procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

26. Neo4j Status Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

27. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

28. License . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260

© 2016 Neo Technology

License: Creative Commons 3.0

IntroductionThis is the developer manual for Neo4j version 3.0, authored by the Neo4j Team.

The main parts of the manual are:

• Introduction — Introducing graph database concepts and Neo4j.

• Get started — Get started using Neo4j: Cypher and Drivers.

• Cypher Query Language — Reference for the Cypher query language.

• Drivers — Uniform language driver manual.

• Appendix — An appendix that covers Neo4j status codes and a Graph database terminology.

Who should read this?

This manual is written for the developer of a Neo4j client application.

1

Chapter 1. Neo4j highlightsConnected data is all around us. Neo4j supports rapid development of graph powered systems thattake advantage of the rich connectedness of data.

A native graph database: Neo4j is built from the ground up to be a graph database. The architectureis designed for optimizing fast management, storage, and traversal of nodes and relationships. InNeo4j, relationships are first class citizens that represent pre-materialized connections betweenentities. An operation known in the relational database world as a join, whose performance degradesexponentially with the number of relationships, is performed by Neo4j as navigation from one node toanother, whose performance is linear.

This different approach to storing and querying connections between entities provides traversalperformance of up to 4 million hops per second and core. As most graph searches are local to thelarger neighborhood of a node, the total amount of data stored in a database will not affectoperations runtime. Dedicated memory management, and highly scalable and memory efficientoperations, contribute to the benefits.

Whiteboard friendly: The property graph approach allows consistent use of the same modelthroughout conception, design, implementation, storage, and visualization of any domain or use case.This allows all business stakeholders to participate throughout the development cycle. With theschema optional model, the domain model can be evolved continuously as requirements change,without penalty of expensive schema changes and migrations.

Cypher, the declarative graph query language, is designed to visually represent graph patterns ofnodes and relationships. This highly capable, yet easily readable, query language is centered aroundthe patterns that express concepts or questions from a specific domain. Cypher can also be extendedfor narrow optimizations for specific use cases.

Supports rapid development: Neo4j supports fast development of graph powered systems. Neo4j’sdevelopment stems from the need to run real-time queries on highly related information; somethingno other database can provide. These unique Neo4j features get you up and running quickly andsustain fast application development for highly scalable applications.

Provides true data safety through ACID transactions: Neo4j uses transactions to guarantee that datais persisted in the case of hardware failure or system crashes.

Designed for business-critical and high-performance operations: Neo4j uses a replicated master-slave cluster setup. It can store hundreds of trillions of entities for the largest datasets imaginablewhile being sensitive to compact storage. Neo4j can be deployed as a scalable, fault-tolerant cluster ofmachines. Due to its high scalability, Neo4j clusters require only tens of machines, not hundreds orthousands, saving on cost and operational complexity. Other features for production applicationsinclude hot-backups and extensive monitoring.

Neo4j’s application is only limited by your imagination.

2

Chapter 2. Graph Database ConceptsThis chapter contains an introduction to the graph data model.

2.1. The Neo4j Graph Database

A graph database stores data in a graph, the most generic of data structures, capable of elegantlyrepresenting any kind of data in a highly accessible way.

For graph database terminology, see Terminology.

Here’s an example graph which we will approach step by step in the following sections:

Person

name = 'Tom Hanks'born = 1956

Movie

title = 'Forrest Gump'released = 1994

ACTED_INroles = ['Forrest']

Person

name = 'Robert Zemeckis'born = 1951

DIRECTED

2.1.1. Nodes

A graph records data in nodes and relationships. Both can have properties. This is

sometimes referred to as the "Property Graph Model".

The fundamental units that form a graph are nodes and relationships. In Neo4j, both nodes andrelationships can contain properties.

Nodes are often used to represent entities, but depending on the domain relationships may be usedfor that purpose as well.

In addition to having properties and relationships, nodes can also be labeled with one or more labels.

The simplest possible graph is a single Node. A Node can have zero or more named values referred toas properties. Let’s start out with one node that has a single property named title:

title = 'Forrest Gump'

The next step is to have multiple nodes. Let’s add two more nodes and one more property on thenode in the previous example:




2.1.2. Relationships

Relationships organize the nodes by connecting them. A relationship connects two nodes — a

start node and an end node. Just like nodes, relationships can have properties.

3

Relationships between nodes are a key part of a graph database. They allow for finding related data.Just like nodes, relationships can have properties.

A relationship connects two nodes, and is guaranteed to have valid start and end nodes.

Relationships organize nodes into arbitrary structures, allowing a graph to resemble a list, a tree, amap, or a compound entity — any of which can be combined into yet more complex, richly inter-connected structures.

Our example graph will make a lot more sense once we add relationships to it:





DIRECTED

Our example uses ACTED_IN and DIRECTED as relationship types. The roles property on the ACTED_INrelationship has an array value with a single item in it.

Below is an ACTED_IN relationship, with the Tom Hanks node as start node and Forrest Gump as end node.




You could also say that the Tom Hanks node has an outgoing relationship, while the Forrest Gump nodehas an incoming relationship.

Relationships are equally well traversed in either direction.

This means that there is no need to add duplicate relationships in the oppositedirection (with regard to traversal or performance).

While relationships always have a direction, you can ignore the direction where it is not useful in yourapplication.

Note that a node can have relationships to itself as well:

name = 'Tom Hanks'born = 1956 KNOWS

The example above would mean that Tom Hanks KNOWS himself.

To further enhance graph traversal all relationships have a relationship type.

Let’s have a look at what can be found by simply following the relationships of a node in our examplegraph:

4





DIRECTED

Table 1. Using relationship direction and type

What we want to know Start from Relationship type Direction

get actors in movie movie node ACTED_IN incoming

get movies with actor person node ACTED_IN outgoing

get directors of movie movie node DIRECTED incoming

get movies directed by person node DIRECTED outgoing

2.1.3. Properties

Both nodes and relationships can have properties.

Properties are named values where the name is a string. The supported property values are:

• Numeric values,

• String values,

• Boolean values,

• Lists of any other type of value.

NULL is not a valid property value.

NULLs can instead be modeled by the absence of a key.

For further details on supported property values, see [property-values-detailed].

2.1.4. Labels

Labels assign roles or types to nodes.

A label is a named graph construct that is used to group nodes into sets; all nodes labeled with thesame label belongs to the same set. Many database queries can work with these sets instead of thewhole graph, making queries easier to write and more efficient to execute. A node may be labeledwith any number of labels, including none, making labels an optional addition to the graph.

Labels are used when defining constraints and adding indexes for properties (see Schema).

An example would be a label named User that you label all your nodes representing users with. Withthat in place, you can ask Neo4j to perform operations only on your user nodes, such as finding allusers with a given name.

However, you can use labels for much more. For instance, since labels can be added and removedduring runtime, they can be used to mark temporary states for your nodes. You might create anOffline label for phones that are offline, a Happy label for happy pets, and so on.

5

In our example, we’ll add Person and Movie labels to our graph:

Person


Movie



Person


DIRECTED

A node can have multiple labels, let’s add an Actor label to the Tom Hanks node.

PersonActor


Label names

Any non-empty Unicode string can be used as a label name. In Cypher, you may need to use thebacktick (`) syntax to avoid clashes with Cypher identifier rules or to allow non-alphanumericcharacters in a label. By convention, labels are written with CamelCase notation, with the first letter inupper case. For instance, User or CarOwner.

Labels have an id space of an int, meaning the maximum number of labels the database can containis roughly 2 billion.

2.1.5. Traversal

A traversal navigates through a graph to find paths.

A traversal is how you query a graph, navigating from starting nodes to related nodes, finding answersto questions like "what music do my friends like that I don’t yet own," or "if this power supply goesdown, what web services are affected?"

Traversing a graph means visiting its nodes, following relationships according to some rules. In mostcases only a subgraph is visited, as you already know where in the graph the interesting nodes andrelationships are found.

Cypher provides a declarative way to query the graph powered by traversals and other techniques.See Cypher Query Language for more information.

If we want to find out which movies Tom Hanks acted in according to our tiny example database, thetraversal would start from the Tom Hanks node, follow any ACTED_IN relationships connected to thenode, and end up with Forrest Gump as the result (see the dashed lines):

Person


Movie



Person


DIRECTED

6

2.1.6. Paths

A path is one or more nodes with connecting relationships, typically retrieved as a query or

traversal result.

In the previous example, the traversal result could be returned as a path:

Person name = 'Tom Hanks'born = 1956 Movie title = 'Forrest Gump'

released = 1994


The path above has length one.

The shortest possible path has length zero — that is, it contains only a single node and norelationships — and can look like this:

Person


This path has length one:

Person


KNOWS

2.1.7. Schema

Neo4j is a schema-optional graph database.

You can use Neo4j without any schema. Optionally, you can introduce it in order to gain performanceor modeling benefits. This allows a way of working where the schema does not get in your way untilyou are at a stage where you want to reap the benefits of having one.

Schema commands can only be applied on the master machine in a Neo4j cluster. Ifyou apply them on a slave you will receive aNeo.ClientError.Transaction.InvalidType error code (see Neo4j Status Codes).

Indexes

Performance is gained by creating indexes, which improve the speed of looking up nodes in

the database.

Once you have specified which properties to index, Neo4j will make sure your indexes are kept up todate as your graph evolves. Any operation that looks up nodes by the newly indexed properties willsee a significant performance boost.

Indexes in Neo4j are eventually available. That means that when you first create an index the operationreturns immediately. The index is populating in the background and so is not immediately available forquerying. When the index has been fully populated it will eventually come online. That means that it isnow ready to be used in queries.

If something should go wrong with the index, it can end up in a failed state. When it is failed, it will notbe used to speed up queries. To rebuild it, you can drop and recreate the index. Look at logs for clues

7

about the failure.

For working with indexes in Cypher, see Indexes

Constraints

Neo4j can help keep your data clean. It does so using constraints. Constraints allow you to specify therules for what your data should look like. Any changes that break these rules will be denied.

For working with constraints in Cypher, see Constraints

Get started

8

Chapter 3. Install Neo4jSetting up a local environment for developing an application with Neo4j and its query languageCypher is easy. Download Neo4j from http://neo4j.com/download/ and follow the installationinstructions for your operating system. Read more about deploying Neo4j in the Neo4j OperationsManual → Deployment (http://neo4j.com/docs/operations-manual/3.0/#deployment). To get started withCypher, continue reading. Open up a web browser and point it to http://localhost:7474 to follow alongand try out the queries explained below.

9

http://neo4j.com/download/

http://neo4j.com/docs/operations-manual/3.0/#deployment

http://neo4j.com/docs/operations-manual/3.0/#deployment

http://localhost:7474

Chapter 4. Get started with CypherThis guide will introduce you to Cypher, Neo4j’s query language. It will help you:

• start thinking about graphs and patterns

• apply this knowledge to simple problems

• learn how to write Cypher statements

4.1. Patterns

Neo4j’s Property Graphs are composed of nodes and relationships, either of which may haveproperties. Nodes represent entities, for example concepts, events, places and things. Relationshipsconnect pairs of nodes.

However, nodes and relationships are simply low-level building blocks. The real strength of theproperty graph lies in its ability to encode patterns of connected nodes and relationships. A singlenode or relationship typically encodes very little information, but a pattern of nodes and relationshipscan encode arbitrarily complex ideas.

Cypher, Neo4j’s query language, is strongly based on patterns. Specifically, patterns are used to matchdesired graph structures. Once a matching structure has been found or created, Neo4j can use it forfurther processing.

A simple pattern, which has only a single relationship, connects a pair of nodes (or, occasionally, anode to itself). For example, a Person LIVES_IN a City or a City is PART_OF a Country.

Complex patterns, using multiple relationships, can express arbitrarily complex concepts and supporta variety of interesting use cases. For example, we might want to match instances where a PersonLIVES_IN a Country. The following Cypher code combines two simple patterns into a (mildly) complexpattern which performs this match:

(:Person) -[:LIVES_IN]-> (:City) -[:PART_OF]-> (:Country)

Pattern recognition is fundamental to the way that the brain works. Consequently, humans are verygood at working with patterns. When patterns are presented visually, for example in a diagram ormap, humans can use them to recognize, specify, and understand concepts. As a pattern-basedlanguage, Cypher takes advantage of this capability.

Like SQL, used in relational databases, Cypher is a textual declarative query language. It uses a form ofASCII art (https://en.wikipedia.org/wiki/ASCII_art) to represent graph-related patterns. SQL-like clauses andkeywords, for example MATCH, WHERE and DELETE are used to combine these patterns and specifydesired actions.

This combination tells Neo4j which patterns to match and what to do with the matching items, forexample nodes, relationships, paths and lists. However, Cypher does not tell Neo4j how to find nodes,traverse relationships etc.

Diagrams made up of icons and arrows are commonly used to visualize graphs. Textual annotationsprovide labels, define properties etc.

4.1.1. Node Syntax

Cypher uses a pair of parentheses (usually containing a text string) to represent a node, eg: (), (foo).This is reminiscent of a circle or a rectangle with rounded end caps. Here are some ASCII-artencodings for example Neo4j nodes, providing varying types and amounts of detail:

10

https://en.wikipedia.org/wiki/ASCII_art

()(matrix)(:Movie)(matrix:Movie)(matrix:Movie {title: "The Matrix"})(matrix:Movie {title: "The Matrix", released: 1997})

The simplest form, (), represents an anonymous, uncharacterized node. If we want to refer to thenode elsewhere, we can add a variable, for example: (matrix). A variable is restricted to a singlestatement. It may have different (or no) meaning in another statement.

The Movie label (prefixed in use with a colon) declares the node’s type. This restricts the pattern,keeping it from matching (say) a structure with an Actor node in this position. Neo4j’s node indexesalso use labels: each index is specific to the combination of a label and a property.

The node’s properties, for example title, are represented as a list of key/value pairs, enclosed withina pair of braces, for example: {name: "Keanu Reeves"}. Properties can be used to store informationand/or restrict patterns.

4.1.2. Relationship Syntax

Cypher uses a pair of dashes (--) to represent an undirected relationship. Directed relationships havean arrowhead at one end (←-, -→). Bracketed expressions ([…]) can be used to add details. This mayinclude variables, properties, and/or type information:

-->-[role]->-[:ACTED_IN]->-[role:ACTED_IN]->-[role:ACTED_IN {roles: ["Neo"]}]->

The syntax and semantics found within a relationship’s bracket pair are very similar to those usedbetween a node’s parentheses. A variable (eg, role) can be defined, to be used elsewhere in thestatement. The relationship’s type (eg, ACTED_IN) is analogous to the node’s label. The properties (eg,roles) are entirely equivalent to node properties. (Note that the value of a property may be an array.)

4.1.3. Pattern Syntax

Combining the syntax for nodes and relationships, we can express patterns. The following could be asimple pattern (or fact) in this domain:

(keanu:Person:Actor {name: "Keanu Reeves"} )-[role:ACTED_IN {roles: ["Neo"] } ]->(matrix:Movie {title: "The Matrix"} )

Like with node labels, the relationship type ACTED_IN is added as a symbol, prefixed with a colon::ACTED_IN. Variables (eg, role) can be used elsewhere in the statement to refer to the relationship.Node and relationship properties use the same notation. In this case, we used an array property forthe roles, allowing multiple roles to be specified.

Pattern Nodes vs. Database Nodes

When a node is used in a pattern, it describes zero or more nodes in the database.Similarly, each pattern describes zero or more paths of nodes and relationships.

11

4.1.4. Pattern Variables

To increase modularity and reduce repetition, Cypher allows patterns to be assigned to variables. Thisallows the matching paths to be inspected, used in other expressions, etc.

acted_in = (:Person)-[:ACTED_IN]->(:Movie)

The acted_in variable would contain two nodes and the connecting relationship for each path thatwas found or created. There are a number of functions to access details of a path, includingnodes(path), rels(path) (same as relationships(path)), and length(path).

4.1.5. Clauses

Cypher statements typically have multiple clauses, each of which performs a specific task, for example:

• create and match patterns in the graph

• filter, project, sort, or paginate results

• compose partial statements

By combining Cypher clauses, we can compose more complex statements that express what we wantto know or create. Neo4j then figures out how to achieve the desired goal in an efficient manner.

4.2. Patterns in Practice

4.2.1. Creating Data

We’ll start by looking into the clauses that allow us to create data.

To add data, we just use the patterns we already know. By providing patterns we can specify whatgraph structures, labels and properties we would like to make part of our graph.

Obviously the simplest clause is called CREATE. It will just go ahead and directly create the patterns thatyou specify.

For the patterns we’ve looked at so far this could look like the following:

CREATE (:Movie { title:"The Matrix",released:1997 })

If we execute this statement, Cypher returns the number of changes, in this case adding 1 node, 1label and 2 properties.

+-------------------+| No data returned. |+-------------------+Nodes created: 1Properties set: 2Labels added: 1

As we started out with an empty database, we now have a database with a single node in it:

Movie

title = 'The Matrix'released = 1997

12

If case we also want to return the created data we can add a RETURN clause, which refers to thevariable we’ve assigned to our pattern elements.

CREATE (p:Person { name:"Keanu Reeves", born:1964 })RETURN p

This is what gets returned:

+----------------------------------------+| p |+----------------------------------------+| Node[1]{name:"Keanu Reeves",born:1964} |+----------------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

If we want to create more than one element, we can separate the elements with commas or usemultiple CREATE statements.

We can of course also create more complex structures, like an ACTED_IN relationship with informationabout the character, or DIRECTED ones for the director.

CREATE (a:Person { name:"Tom Hanks", born:1956 })-[r:ACTED_IN { roles: ["Forrest"]}]->(m:Movie { title:"Forrest Gump",released:1994 })CREATE (d:Person { name:"Robert Zemeckis", born:1951 })-[:DIRECTED]->(m)RETURN a,d,r,m

This is the part of the graph we just updated:

Person


Movie



Person


DIRECTED

In most cases, we want to connect new data to existing structures. This requires that we know how tofind existing patterns in our graph data, which we will look at next.

4.2.2. Matching Patterns

Matching patterns is a task for the MATCH statement. We pass the same kind of patterns we’ve used sofar to MATCH to describe what we’re looking for. It is similar to query by example, only that our examplesalso include the structures.

A MATCH statement will search for the patterns we specify and return one row persuccessful pattern match.

To find the data we’ve created so far, we can start looking for all nodes labeled with the Movie label.

MATCH (m:Movie)RETURN m

13

Here’s the result:

This should show both The Matrix and Forrest Gump.

We can also look for a specific person, like Keanu Reeves.

MATCH (p:Person { name:"Keanu Reeves" })RETURN p

This query returns the matching node:

Note that we only provide enough information to find the nodes, not all properties are required. Inmost cases you have key-properties like SSN, ISBN, emails, logins, geolocation or product codes tolook for.

We can also find more interesting connections, like for instance the movies titles that Tom Hanks actedin and the roles he played.

MATCH (p:Person { name:"Tom Hanks" })-[r:ACTED_IN]->(m:Movie)RETURN m.title, r.roles

+------------------------------+| m.title | r.roles |+------------------------------+| "Forrest Gump" | ["Forrest"] |+------------------------------+1 row

In this case we only returned the properties of the nodes and relationships that we were interested in.You can access them everywhere via a dot notation identifer.property.

Of course this only lists his role as Forrest in Forrest Gump because that’s all data that we’ve added.

Now we know enough to connect new nodes to existing ones and can combine MATCH and CREATE toattach structures to the graph.

4.2.3. Attaching Structures

To extend the graph with new information, we first match the existing connection points and thenattach the newly created nodes to them with relationships. Adding Cloud Atlas as a new movie for TomHanks could be achieved like this:

MATCH (p:Person { name:"Tom Hanks" })CREATE (m:Movie { title:"Cloud Atlas",released:2012 })CREATE (p)-[r:ACTED_IN { roles: ['Zachry']}]->(m)RETURN p,r,m

Here’s what the structure looks like in the database:

14

Person


Movie

title = 'Cloud Atlas'released = 2012

ACTED_INroles = ['Zachry']

It is important to remember that we can assign variables to both nodes andrelationships and use them later on, no matter if they were created or matched.

It is possible to attach both node and relationship in a single CREATE clause. For readability it helps tosplit them up though.

A tricky aspect of the combination of MATCH and CREATE is that we get one row permatched pattern. This causes subsequent CREATE statements to be executed once foreach row. In many cases this is what you want. If that’s not intended, please movethe CREATE statement before the MATCH, or change the cardinality of the query withmeans discussed later or use the get or create semantics of the next clause: MERGE.

4.2.4. Completing Patterns

Whenever we get data from external systems or are not sure if certain information already exists inthe graph, we want to be able to express a repeatable (idempotent) update operation. In Cypher MERGEhas this function. It acts like a combination of MATCH or CREATE, which checks for the existence of datafirst before creating it. With MERGE you define a pattern to be found or created. Usually, as with MATCHyou only want to include the key property to look for in your core pattern. MERGE allows you to provideadditional properties you want to set ON CREATE.

If we wouldn’t know if our graph already contained Cloud Atlas we could merge it in again.

MERGE (m:Movie { title:"Cloud Atlas" })ON CREATE SET m.released = 2012RETURN m

+--------------------------------------------+| m |+--------------------------------------------+| Node[5]{title:"Cloud Atlas",released:2012} |+--------------------------------------------+1 row

We get a result in any both cases: either the data (potentially more than one row) that was already inthe graph or a single, newly created Movie node.

A MERGE clause without any previously assigned variables in it either matches the fullpattern or creates the full pattern. It never produces a partial mix of matching andcreating within a pattern. To achieve a partial match/create, make sure to usealready defined variables for the parts that shouldn’t be affected.

So foremost MERGE makes sure that you can’t create duplicate information or structures, but it comeswith the cost of needing to check for existing matches first. Especially on large graphs it can be costlyto scan a large set of labeled nodes for a certain property. You can alleviate some of that by creatingsupporting indexes or constraints, which we’ll discuss later. But it’s still not for free, so wheneveryou’re sure to not create duplicate data use CREATE over MERGE.

15

MERGE can also assert that a relationship is only created once. For that to work youhave to pass in both nodes from a previous pattern match.

MATCH (m:Movie { title:"Cloud Atlas" })MATCH (p:Person { name:"Tom Hanks" })MERGE (p)-[r:ACTED_IN]->(m)ON CREATE SET r.roles =['Zachry']RETURN p,r,m

Person


Movie



In case the direction of a relationship is arbitrary, you can leave off the arrowhead. MERGE will thencheck for the relationship in either direction, and create a new directed relationship if no matchingrelationship was found.

If you choose to pass in only one node from a preceding clause, MERGE offers an interestingfunctionality. It will then only match within the direct neighborhood of the provided node for the givenpattern, and, if not found create it. This can come in very handy for creating for example treestructures.

CREATE (y:Year { year:2014 })MERGE (y)<-[:IN_YEAR]-(m10:Month { month:10 })MERGE (y)<-[:IN_YEAR]-(m11:Month { month:11 })RETURN y,m10,m11

This is the graph structure that gets created:

Year

year = 2014

Month

month = 11

IN_YEAR

Month

month = 10

IN_YEAR

Here there is no global search for the two Month nodes; they are only searched for in the context of the2014 Year node.

4.3. Getting correct results

Let’s first get some data in to retrieve results from:

16

CREATE (matrix:Movie { title:"The Matrix",released:1997 })CREATE (cloudAtlas:Movie { title:"Cloud Atlas",released:2012 })CREATE (forrestGump:Movie { title:"Forrest Gump",released:1994 })CREATE (keanu:Person { name:"Keanu Reeves", born:1964 })CREATE (robert:Person { name:"Robert Zemeckis", born:1951 })CREATE (tom:Person { name:"Tom Hanks", born:1956 })CREATE (tom)-[:ACTED_IN { roles: ["Forrest"]}]->(forrestGump)CREATE (tom)-[:ACTED_IN { roles: ['Zachry']}]->(cloudAtlas)CREATE (robert)-[:DIRECTED]->(forrestGump)

This is the data we will start out with:

Movie

title = 'The Matrix'released = 1997

Movie


Movie


Person

name = 'Keanu Reeves'born = 1964

Person


DIRECTED

Person




4.3.1. Filtering Results

So far we’ve matched patterns in the graph and always returned all results we found. Quite oftenthere are conditions in play for what we want to see. Similar to in SQL those filter conditions areexpressed in a WHERE clause. This clause allows to use any number of boolean expressions (predicates)combined with AND, OR, XOR and NOT. The simplest predicates are comparisons, especially equality.

MATCH (m:Movie)WHERE m.title = "The Matrix"RETURN m

+-------------------------------------------+| m |+-------------------------------------------+| Node[0]{title:"The Matrix",released:1997} |+-------------------------------------------+1 row

For equality on one or more properties, a more compact syntax can be used as well:

MATCH (m:Movie { title: "The Matrix" })RETURN m

Other options are numeric comparisons, matching regular expressions and checking the existence ofvalues within a list.

The WHERE clause below includes a regular expression match, a greater than comparison and a test tosee if a value exists in a list.

MATCH (p:Person)-[r:ACTED_IN]->(m:Movie)WHERE p.name =~ "K.+" OR m.released > 2000 OR "Neo" IN r.rolesRETURN p,r,m

17

+-------------------------------------------------------------------------------------------------------------------+| p | r | m|+-------------------------------------------------------------------------------------------------------------------+| Node[5]{name:"Tom Hanks",born:1956} | :ACTED_IN[1]{roles:["Zachry"]} | Node[1]{title:"CloudAtlas",released:2012} |+-------------------------------------------------------------------------------------------------------------------+1 row

One aspect that might be a little surprising is that you can even use patterns as predicates. WhereMATCH expands the number and shape of patterns matched, a pattern predicate restricts the currentresult set. It only allows the paths to pass that satisfy the additional patterns as well (or NOT).

MATCH (p:Person)-[:ACTED_IN]->(m)WHERE NOT (p)-[:DIRECTED]->()RETURN p,m

+-----------------------------------------------------------------------------------+| p | m |+-----------------------------------------------------------------------------------+| Node[5]{name:"Tom Hanks",born:1956} | Node[1]{title:"Cloud Atlas",released:2012} || Node[5]{name:"Tom Hanks",born:1956} | Node[2]{title:"Forrest Gump",released:1994} |+-----------------------------------------------------------------------------------+2 rows

Here we find actors, because they sport an ACTED_IN relationship but then skip those that everDIRECTED any movie.

There are also more advanced ways of filtering like list-predicates which we will look at later on.

4.3.2. Returning Results

So far we’ve returned only nodes, relationships, or paths directly via their variables. But the RETURNclause can actually return any number of expressions. But what are actually expressions in Cypher?

The simplest expressions are literal values like numbers, strings and arrays as [1,2,3], and maps like{name:"Tom Hanks", born:1964, movies:["Forrest Gump", …], count:13}. You can access individualproperties of any node, relationship, or map with a dot-syntax like n.name. Individual elements orslices of arrays can be retrieved with subscripts like names[0] or movies[1..-1]. Each functionevaluation like length(array), toInt("12"), substring("2014-07-01",0,4), orcoalesce(p.nickname,"n/a") is also an expression.

Predicates that you’d use in WHERE count as boolean expressions.

Of course simpler expressions can be composed and concatenated to form more complexexpressions.

By default the expression itself will be used as label for the column, in many cases you want to aliasthat with a more understandable name using expression AS alias. You can later on refer to thatcolumn using its alias.

MATCH (p:Person)RETURN p, p.name AS name, upper(p.name), coalesce(p.nickname,"n/a") AS nickname, { name: p.name, label:head(labels(p))} AS person

18

+-----------------------------------------------------------------------------------------------------------------------------------------------+| p | name | upper(p.name) | nickname | person|+-----------------------------------------------------------------------------------------------------------------------------------------------+| Node[3]{name:"Keanu Reeves",born:1964} | "Keanu Reeves" | "KEANU REEVES" | "n/a" | {name ->"Keanu Reeves", label -> "Person"} || Node[4]{name:"Robert Zemeckis",born:1951} | "Robert Zemeckis" | "ROBERT ZEMECKIS" | "n/a" | {name ->"Robert Zemeckis", label -> "Person"} || Node[5]{name:"Tom Hanks",born:1956} | "Tom Hanks" | "TOM HANKS" | "n/a" | {name ->"Tom Hanks", label -> "Person"} |+-----------------------------------------------------------------------------------------------------------------------------------------------+3 rows

If you’re interested in unique results you can use the DISTINCT keyword after RETURN to indicate that.

4.3.3. Aggregating Information

In many cases you want to aggregate or group the data that you encounter while traversing patternsin your graph. In Cypher aggregation happens in the RETURN clause while computing your final results.Many common aggregation functions are supported, e.g. count, sum, avg, min, and max, but there areseveral more.

Counting the number of people in your database could be achieved by this:

MATCH (:Person)RETURN count(*) AS people

+--------+| people |+--------+| 3 |+--------+1 row

Please note that NULL values are skipped during aggregation. For aggregating only unique values useDISTINCT, like in count(DISTINCT role).

Aggregation in Cypher just works. You specify which result columns you want to aggregate and Cypherwill use all non-aggregated columns as grouping keys.

Aggregation affects which data is still visible in ordering or later query parts.

To find out how often an actor and director worked together, you’d run this statement:

MATCH (actor:Person)-[:ACTED_IN]->(movie:Movie)<-[:DIRECTED]-(director:Person)RETURN actor,director,count(*) AS collaborations

+--------------------------------------------------------------------------------------------------+| actor | director | collaborations |+--------------------------------------------------------------------------------------------------+| Node[5]{name:"Tom Hanks",born:1956} | Node[4]{name:"Robert Zemeckis",born:1951} | 1 |+--------------------------------------------------------------------------------------------------+1 row

Frequently you want to sort and paginate after aggregating a count(x).

19

4.3.4. Ordering and Pagination

Ordering works like in other query languages, with an ORDER BY expression [ASC|DESC] clause. Theexpression can be any expression discussed before as long as it is computable from the returnedinformation.

So for instance if you return person.name you can still ORDER BY person.age as both are accessible fromthe person reference. You cannot order by things that you can’t infer from the information you return.This is especially important with aggregation and DISTINCT return values as both remove the visibilityof data that is aggregated.

Pagination is a straightforward use of SKIP {offset} LIMIT {count}.

A common pattern is to aggregate for a count (score or frequency), order by it and only return the top-n entries.

For instance to find the most prolific actors you could do:

MATCH (a:Person)-[:ACTED_IN]->(m:Movie)RETURN a,count(*) AS appearancesORDER BY appearances DESC LIMIT 10;

+---------------------------------------------------+| a | appearances |+---------------------------------------------------+| Node[5]{name:"Tom Hanks",born:1956} | 2 |+---------------------------------------------------+1 row

4.3.5. Collecting Aggregation

The most helpful aggregation function is collect(), which, appropriately collects all aggregated valuesinto a list. This comes very handy in many situations as no information of details is lost whileaggregating.

collect() is well suited for retrieving typical parent-child structures, where one core entity (parent,root or head) is returned per row with all its dependent information in associated lists created withcollect(). This means that there is no need to repeat the parent information per each child-row, oreven running n+1 statements to retrieve the parent and its children individually.

To retrieve the cast of each movie in our database this statement could be used:

MATCH (m:Movie)<-[:ACTED_IN]-(a:Person)RETURN m.title AS movie, collect(a.name) AS cast, count(*) AS actors

+-----------------------------------------+| movie | cast | actors |+-----------------------------------------+| "Forrest Gump" | ["Tom Hanks"] | 1 || "Cloud Atlas" | ["Tom Hanks"] | 1 |+-----------------------------------------+2 rows

The lists created by collect() can either be used from the client consuming the Cypher results ordirectly within a statement with any of the list functions or predicates.

20

4.4. Composing large statements

Let’s first get some data in to retrieve results from:

CREATE (matrix:Movie { title:"The Matrix",released:1997 })CREATE (cloudAtlas:Movie { title:"Cloud Atlas",released:2012 })CREATE (forrestGump:Movie { title:"Forrest Gump",released:1994 })CREATE (keanu:Person { name:"Keanu Reeves", born:1964 })CREATE (robert:Person { name:"Robert Zemeckis", born:1951 })CREATE (tom:Person { name:"Tom Hanks", born:1956 })CREATE (tom)-[:ACTED_IN { roles: ["Forrest"]}]->(forrestGump)CREATE (tom)-[:ACTED_IN { roles: ['Zachry']}]->(cloudAtlas)CREATE (robert)-[:DIRECTED]->(forrestGump)

4.4.1. UNION

A Cypher statement is usually quite compact. Expressing references between nodes as visual patternsmakes them easy to understand.

If you want to combine the results of two statements that have the same result structure, you can useUNION [ALL].

For instance if you want to list both actors and directors without using the alternative relationship-type syntax ()-[:ACTED_IN|:DIRECTED]→() you can do this:

MATCH (actor:Person)-[r:ACTED_IN]->(movie:Movie)RETURN actor.name AS name, type(r) AS acted_in, movie.title AS titleUNIONMATCH (director:Person)-[r:DIRECTED]->(movie:Movie)RETURN director.name AS name, type(r) AS acted_in, movie.title AS title

+-------------------------------------------------+| name | acted_in | title |+-------------------------------------------------+| "Tom Hanks" | "ACTED_IN" | "Cloud Atlas" || "Tom Hanks" | "ACTED_IN" | "Forrest Gump" || "Robert Zemeckis" | "DIRECTED" | "Forrest Gump" |+-------------------------------------------------+3 rows

4.4.2. WITH

In Cypher it’s possible to chain fragments of statements together, much like you would do within adata-flow pipeline. Each fragment works on the output from the previous one and its results can feedinto the next one.

You use the WITH clause to combine the individual parts and declare which data flows from one to theother. WITH is very much like RETURN with the difference that it doesn’t finish a query but prepares theinput for the next part. You can use the same expressions, aggregations, ordering and pagination as inthe RETURN clause.

The only difference is that you must alias all columns as they would otherwise not be accessible. Onlycolumns that you declare in your WITH clause is available in subsequent query parts.

See below for an example where we collect the movies someone appeared in, and then filter outthose which appear in only one movie.

21

MATCH (person:Person)-[:ACTED_IN]->(m:Movie)WITH person, count(*) AS appearances, collect(m.title) AS moviesWHERE appearances > 1RETURN person.name, appearances, movies

+------------------------------------------------------------+| person.name | appearances | movies |+------------------------------------------------------------+| "Tom Hanks" | 2 | ["Cloud Atlas","Forrest Gump"] |+------------------------------------------------------------+1 row

If you want to filter by an aggregated value in SQL or similar languages you wouldhave to use HAVING. That’s a single purpose clause for filtering aggregatedinformation. In Cypher, WHERE can be used in both cases.

4.5. Constraints and indexes

Labels are a convenient way to group nodes together. They are used to restrict queries, defineconstraints and create indexes.

4.5.1. Using constraints

You can specify unique constraints that guarantee uniqueness of a certain property on nodes with aspecific label. These constraints are also used by the MERGE clause to make certain that a node onlyexists once.

The following is an example of how to use labels and add constraints and indexes to them. Let’s startout by adding a constraint. In this case we decide that every Movie node should have a unique title.

CREATE CONSTRAINT ON (movie:Movie) ASSERT movie.title IS UNIQUE

Note that adding the unique constraint will implicitly add an index on that property, so we won’t haveto do that separately. If we drop a constraint but still want an index on that property, we will have tocreate the index explicitly.

Constraints can be added after a label is already in use, but that requires that the existing datacomplies with the constraints.

4.5.2. Using indexes

The main reason for using indexes in a graph database is to find the starting point in the graph as fastas possible. After that seek you rely on in-graph structures and the first class citizenship ofrelationships in the graph database to achieve high performance. Thus graph queries themselves donot need indexes to run fast.

Indexes can be added at any time. Note that it will take some time for an index to come online whenthere’s existing data.

In this case we want to create an index to speed up finding actors by name in the database:

CREATE INDEX ON :Actor(name)

Now, let’s add some data.

22

CREATE (actor:Actor { name:"Tom Hanks" }),(movie:Movie { title:'Sleepless IN Seattle' }), (actor)-[:ACTED_IN]->(movie);

Normally you don’t specify indexes when querying for data. They will be used automatically. Thismeans we that can simply look up the Tom Hanks node, and the index will kick in behind the scenes toboost performance.

MATCH (actor:Actor { name: "Tom Hanks" })RETURN actor;

Cypher Query LanguageThe Cypher part is the authoritative source for details on the Cypher Query Language. For a shortintroduction, see What is Cypher?. To take your first steps with Cypher, see Get started with Cypher.For the terminology used, see Terminology.

23

Chapter 5. IntroductionTo get an overview of Cypher, continue reading What is Cypher?. The rest of this chapter deals withthe context of Cypher statements, like for example transaction management and how to useparameters. For the Cypher language reference itself see other chapters at Cypher Query Language.To take your first steps with Cypher, see Get started with Cypher. For the terminology used, seeTerminology.

5.1. What is Cypher?

5.1.1. Introduction

Cypher is a declarative graph query language that allows for expressive and efficient querying andupdating of the graph store. Cypher is a relatively simple but still very powerful language. Verycomplicated database queries can easily be expressed through Cypher. This allows you to focus onyour domain instead of getting lost in database access.

Cypher is designed to be a humane query language, suitable for both developers and (importantly, wethink) operations professionals. Our guiding goal is to make the simple things easy, and the complexthings possible. Its constructs are based on English prose and neat iconography which helps to makequeries more self-explanatory. We have tried to optimize the language for reading and not for writing.

Being a declarative language, Cypher focuses on the clarity of expressing what to retrieve from agraph, not on how to retrieve it. This is in contrast to imperative languages like Java, scriptinglanguages like Gremlin (http://gremlin.tinkerpop.com), and the JRuby Neo4j bindings(https://github.com/neo4jrb/neo4j/). This approach makes query optimization an implementation detailinstead of burdening the user with it and requiring her to update all traversals just because thephysical database structure has changed (new indexes etc.).

Cypher is inspired by a number of different approaches and builds upon established practices forexpressive querying. Most of the keywords like WHERE and ORDER BY are inspired by SQL(http://en.wikipedia.org/wiki/SQL). Pattern matching borrows expression approaches from SPARQL(http://en.wikipedia.org/wiki/SPARQL). Some of the collection semantics have been borrowed fromlanguages such as Haskell and Python.

5.1.2. Structure

Cypher borrows its structure from SQL — queries are built up using various clauses.

Clauses are chained together, and the they feed intermediate result sets between each other. Forexample, the matching variables from one MATCH clause will be the context that the next clause existsin.

The query language is comprised of several distinct clauses. You can read more details about themlater in the manual.

Here are a few clauses used to read from the graph:

• MATCH: The graph pattern to match. This is the most common way to get data from the graph.

• WHERE: Not a clause in its own right, but rather part of MATCH, OPTIONAL MATCH and WITH. Addsconstraints to a pattern, or filters the intermediate result passing through WITH.

• RETURN: What to return.

Let’s see MATCH and RETURN in action.

Imagine an example graph like the following one:

24

http://gremlin.tinkerpop.com

https://github.com/neo4jrb/neo4j/

http://en.wikipedia.org/wiki/SQL

http://en.wikipedia.org/wiki/SPARQL

name = 'Joe'

name = 'Steve'

friend

name = 'John'

friend

name = 'Sara'

friend

name = 'Maria'

friend

Figure 1. Example Graph

For example, here is a query which finds a user called John and John’s friends (though not his directfriends) before returning both John and any friends-of-friends that are found.

MATCH (john {name: 'John'})-[:friend]->()-[:friend]->(fof)RETURN john.name, fof.name

Resulting in:

+----------------------+| john.name | fof.name |+----------------------+| "John" | "Steve" || "John" | "Maria" |+----------------------+2 rows

Next up we will add filtering to set more parts in motion:

We take a list of user names and find all nodes with names from this list, match their friends andreturn only those followed users who have a name property starting with S.

MATCH (user)-[:friend]->(follower)WHERE user.name IN ['Joe', 'John', 'Sara', 'Maria', 'Steve'] AND follower.name =~ 'S.*'RETURN user.name, follower.name

Resulting in:

+---------------------------+| user.name | follower.name |+---------------------------+| "Joe" | "Steve" || "John" | "Sara" |+---------------------------+2 rows

And here are examples of clauses that are used to update the graph:

• CREATE (and DELETE): Create (and delete) nodes and relationships.

• SET (and REMOVE): Set values to properties and add labels on nodes using SET and use REMOVEto remove them.

• MERGE: Match existing or create new nodes and patterns. This is especially useful together withuniqueness constraints.

25

For more Cypher examples, see [data-modeling-examples] as well as the rest of the Cypher part withdetails on the language. To use Cypher from Java, see [tutorials-cypher-java]. To take your first stepswith Cypher, see Get started with Cypher.

5.2. Updating the graph

Cypher can be used for both querying and updating your graph.

5.2.1. The Structure of Updating Queries

A Cypher query part can’t both match and update the graph at the same time.

Every part can either read and match on the graph, or make updates on it.

If you read from the graph and then update the graph, your query implicitly has two parts — thereading is the first part, and the writing is the second part.

If your query only performs reads, Cypher will be lazy and not actually match the pattern until you askfor the results. In an updating query, the semantics are that all the reading will be done before anywriting actually happens.

The only pattern where the query parts are implicit is when you first read and then write — any otherorder and you have to be explicit about your query parts. The parts are separated using the WITHstatement. WITH is like an event horizon — it’s a barrier between a plan and the finished execution ofthat plan.

When you want to filter using aggregated data, you have to chain together two reading queryparts — the first one does the aggregating, and the second filters on the results coming from the firstone.

MATCH (n {name: 'John'})-[:FRIEND]-(friend)WITH n, count(friend) as friendsCountWHERE friendsCount > 3RETURN n, friendsCount

Using WITH, you specify how you want the aggregation to happen, and that the aggregation has to befinished before Cypher can start filtering.

Here’s an example of updating the graph, writing the aggregated data to the graph:

MATCH (n {name: 'John'})-[:FRIEND]-(friend)WITH n, count(friend) as friendsCountSET n.friendCount = friendsCountRETURN n.friendsCount

You can chain together as many query parts as the available memory permits.

5.2.2. Returning data

Any query can return data. If your query only reads, it has to return data — it serves no purpose if itdoesn’t, and it is not a valid Cypher query. Queries that update the graph don’t have to returnanything, but they can.

After all the parts of the query comes one final RETURN clause. RETURN is not part of any query part — itis a period symbol at the end of a query. The RETURN clause has three sub-clauses that come with it:SKIP/LIMIT and ORDER BY.

If you return graph elements from a query that has just deleted them — beware, you are holding a

26

pointer that is no longer valid. Operations on that node are undefined.

5.3. Transactions

Any query that updates the graph will run in a transaction. An updating query will always either fullysucceed, or not succeed at all.

Cypher will either create a new transaction or run inside an existing one:

• If no transaction exists in the running context Cypher will create one and commit it once the queryfinishes.

• In case there already exists a transaction in the running context, the query will run inside it, andnothing will be persisted to disk until that transaction is successfully committed.

This can be used to have multiple queries be committed as a single transaction:

1. Open a transaction,

2. run multiple updating Cypher queries,

3. and commit all of them in one go.

Note that a query will hold the changes in memory until the whole query has finished executing. Alarge query will consequently need a JVM with lots of heap space.

For using transactions over the REST API, see Transactional Cypher HTTP endpoint.

When writing server extensions or using Neo4j embedded, remember that all iterators returned froman execution result should be either fully exhausted or closed to ensure that the resources bound tothem will be properly released. Resources include transactions started by the query, so failing to do somay, for example, lead to deadlocks or other weird behavior.

5.4. Uniqueness

While pattern matching, Neo4j makes sure to not include matches where the same graph relationshipis found multiple times in a single pattern. In most use cases, this is a sensible thing to do.

Example: looking for a user’s friends of friends should not return said user.

Let’s create a few nodes and relationships:

CREATE (adam:User { name: 'Adam' }),(pernilla:User { name: 'Pernilla' }),(david:User { name: 'David' }), (adam)-[:FRIEND]->(pernilla),(pernilla)-[:FRIEND]->(david)

Which gives us the following graph:

27

User

name = 'Adam'

User

name = 'Pernilla'

FRIEND

User

name = 'David'

FRIEND

Now let’s look for friends of friends of Adam:

MATCH (user:User { name: 'Adam' })-[r1:FRIEND]-()-[r2:FRIEND]-(friend_of_a_friend)RETURN friend_of_a_friend.name AS fofName

+---------+| fofName |+---------+| "David" |+---------+1 row

In this query, Cypher makes sure to not return matches where the pattern relationships r1 and r2point to the same graph relationship.

This is however not always desired. If the query should return the user, it is possible to spread thematching over multiple MATCH clauses, like so:

MATCH (user:User { name: 'Adam' })-[r1:FRIEND]-(friend)MATCH (friend)-[r2:FRIEND]-(friend_of_a_friend)RETURN friend_of_a_friend.name AS fofName

+---------+| fofName |+---------+| "David" || "Adam" |+---------+2 rows

Note that while the following query looks similar to the previous one, it is actually equivalent to theone before.

MATCH (user:User { name: 'Adam' })-[r1:FRIEND]-(friend),(friend)-[r2:FRIEND]-(friend_of_a_friend)RETURN friend_of_a_friend.name AS fofName

Here, the MATCH clause has a single pattern with two paths, while the previous query has two distinctpatterns.

+---------+| fofName |+---------+| "David" |+---------+1 row

28

5.5. Parameters

Cypher supports querying with parameters. This means developers don’t have to resort to stringbuilding to create a query. In addition to that, it also makes caching of execution plans much easier forCypher.

Parameters can be used for literals and expressions in the WHERE clause, for the index value in theSTART clause, index queries, and finally for node/relationship ids. Parameters can not be used as forproperty names, relationship types and labels, since these patterns are part of the query structurethat is compiled into a query plan.

Accepted names for parameters are letters and numbers, and any combination of these.

For details on using parameters via the Neo4j REST API, see Transactional Cypher HTTP endpoint. Fordetails on parameters when using the Neo4j embedded Java API, see [tutorials-cypher-parameters-java].

Below follows a comprehensive set of examples of parameter usage. The parameters are given asJSON here. Exactly how to submit them depends on the driver in use.

5.5.1. String literal

Parameters

{ "name" : "Johan"}

Query

MATCH (n)WHERE n.name = { name }RETURN n

You can use parameters in this syntax as well:

Parameters

{ "name" : "Johan"}

Query

MATCH (n { name: { name }})RETURN n

5.5.2. Regular expression

Parameters

{ "regex" : ".*h.*"}

29

Query

MATCH (n)WHERE n.name =~ { regex }RETURN n.name

5.5.3. Case-sensitive string pattern matching

Parameters

{ "name" : "Michael"}

Query

MATCH (n)WHERE n.name STARTS WITH { name }RETURN n.name

5.5.4. Create node with properties

Parameters

{ "props" : { "name" : "Andres", "position" : "Developer" }}

Query

CREATE ({ props })

5.5.5. Create multiple nodes with properties

Parameters

{ "props" : [ { "awesome" : true, "name" : "Andres", "position" : "Developer" }, { "children" : 3, "name" : "Michael", "position" : "Developer" } ]}

Query

UNWIND { props } AS propertiesCREATE (n:Person)SET n = propertiesRETURN n

30

5.5.6. Setting all properties on node

Note that this will replace all the current properties.

Parameters


Query

MATCH (n)WHERE n.name='Michaela'SET n = { props }

5.5.7. SKIP and LIMIT

Parameters

{ "s" : 1, "l" : 1}

Query

MATCH (n)RETURN n.nameSKIP { s }LIMIT { l }

5.5.8. Node id

Parameters

{ "id" : 0}

Query

MATCH (n)WHERE id(n)= { id }RETURN n.name

5.5.9. Multiple node ids

Parameters

{ "ids" : [ 0, 1, 2 ]}

31

Query

MATCH (n)WHERE id(n) IN { ids }RETURN n.name

5.5.10. Index value (legacy indexes)

Parameters

{ "value" : "Michaela"}

Query

START n=node:people(name = { value })RETURN n

5.5.11. Index query (legacy indexes)

Parameters

{ "query" : "name:Andreas"}

Query

START n=node:people({ query })RETURN n

5.6. Compatibility

Cypher is still changing rather rapidly. Parts of the changes are internal — we add new patternmatchers, aggregators and optimizations or write new query planners, which hopefully makes yourqueries run faster.

Other changes are directly visible to our users — the syntax is still changing. New concepts are beingadded and old ones changed to fit into new possibilities. To guard you from having to keep up withour syntax changes, Neo4j allows you to use an older parser, but still gain speed from newoptimizations.

There are two ways you can select which parser to use. You can configure your database with theconfiguration parameter cypher.default_language_version, and enter which parser you’d like to use(see Supported Language Versions)). Any Cypher query that doesn’t explicitly say anything else, will getthe parser you have configured, or the latest parser if none is configured.

The other way is on a query by query basis. By simply putting CYPHER 2.2 at the beginning, thatparticular query will be parsed with the 2.2 version of the parser. Below is an example using the STARTclause to access a legacy index:

CYPHER 2.2START n=node:nodes(name = "A")RETURN n

32

5.6.1. Accessing entities by id via START

In versions of Cypher prior to 2.2 it was also possible to access specific nodes or relationships usingthe START clause. In this case you could use a syntax like the following:

CYPHER 1.9START n=node(42)RETURN n

The use of the START clause to find nodes by ID was deprecated from Cypher 2.0onwards and is now entirely disabled in Cypher 2.2 and up. You should insteadmake use of the MATCH clause for starting points. See Match for more information onthe correct syntax for this. The START clause should only be used when accessinglegacy indexes (see [indexing]).

5.6.2. Supported Language Versions

Neo4j 2.3 supports the following versions of the Cypher language:

• Neo4j Cypher 2.3



Each release of Neo4j supports a limited number of old Cypher Language Versions.When you upgrade to a new release of Neo4j, please make sure that it supports theCypher language version you need. If not, you may need to modify your queries towork with a newer Cypher language version.

33

Chapter 6. SyntaxThe nitty-gritty details of Cypher syntax.

6.1. Values

All values that are handled by Cypher have a distinct type. The supported types of values are:

• Numeric values,

• String values,

• Boolean values,

• Nodes,

• Relationships,

• Paths,

• Maps from Strings to other values,

• Lists of any other type of value.

Most types of values can be constructed in a query using literal expressions (see Expressions). Specialcare must be taken when using NULL, as NULL is a value of every type (see Working with NULL). Nodes,relationships, and paths are returned as a result of pattern matching.

Note that labels are not values but are a form of pattern syntax.

6.2. Expressions

6.2.1. Expressions in general

An expression in Cypher can be:

• A decimal (integer or double) literal: 13, -40000, 3.14, 6.022E23.

• A hexadecimal integer literal (starting with 0x): 0x13zf, 0xFC3A9, -0x66eff.

• An octal integer literal (starting with 0): 01372, 02127, -05671.

• A string literal: "Hello", 'World'.

• A boolean literal: true, false, TRUE, FALSE.

• A variable: n, x, rel, myFancyVariable, \`A name with weird stuff in it[]!`.

• A property: n.prop, x.prop, rel.thisProperty, myFancyVariable.\`(weird property name)`.

• A dynamic property: n["prop"], rel[n.city + n.zip], map[coll[0]].

• A parameter: {param}, {0}

• A list of expressions: ["a", "b"], [1,2,3], ["a", 2, n.property, {param}], [ ].

• A function call: length(p), nodes(p).

• An aggregate function: avg(x.prop), count(*).

• A path-pattern: (a)-→()←-(b).

• An operator application: 1 + 2 and 3 < 4.

• A predicate expression is an expression that returns true or false: a.prop = "Hello", length(p) >10, has(a.name).

34

• A regular expression: a.name =~ "Tob.*"

• A case-sensitive string matching expression: a.surname STARTS WITH "Sven", a.surname ENDS WITH"son" or a.surname CONTAINS "son"

• A CASE expression.

6.2.2. Note on string literals

String literals can contain these escape sequences.

Escape sequence Character

\t Tab

\b Backspace

\n Newline

\r Carriage return

\f Form feed

\' Single quote

\" Double quote

\\ Backslash

\uxxxx Unicode UTF-16 code point (4 hexdigits must follow the \u)

\Uxxxxxxxx Unicode UTF-32 code point (8 hexdigits must follow the \U)

6.2.3. Case Expressions

Cypher supports CASE expressions, which is a generic conditional expression, similar to if/elsestatements in other languages. Two variants of CASE exist — the simple form and the generic form.

6.2.4. Simple CASE

The expression is calculated, and compared in order with the WHEN clauses until a match is found. Ifno match is found the expression in the ELSE clause is used, or null, if no ELSE case exists.

Syntax:

CASE testWHEN value THEN result[WHEN ...][ELSE default]END

Arguments:

• test: A valid expression.

• value: An expression whose result will be compared to the test expression.

• result: This is the result expression used if the value expression matches the test expression.

• default: The expression to use if no match is found.

35

Query

MATCH (n)RETURNCASE n.eyesWHEN 'blue'THEN 1WHEN 'brown'THEN 2ELSE 3 END AS result

Result

+--------+| result |+--------+| 2 || 1 || 3 || 2 || 1 |+--------+5 rows

6.2.5. Generic CASE

The predicates are evaluated in order until a true value is found, and the result value is used. If nomatch is found the expression in the ELSE clause is used, or null, if no ELSE case exists.

Syntax:

CASEWHEN predicate THEN result[WHEN ...][ELSE default]END

Arguments:

• predicate: A predicate that is tested to find a valid alternative.

• result: This is the result expression used if the predicate matches.

• default: The expression to use if no match is found.

Query

MATCH (n)RETURNCASEWHEN n.eyes = 'blue'THEN 1WHEN n.age < 40THEN 2ELSE 3 END AS result

36

Result

+--------+| result |+--------+| 2 || 1 || 3 || 3 || 1 |+--------+5 rows

6.3. Variables

When you reference parts of a pattern or a query, you do so by naming them. The names you give thedifferent parts are called variables.

In this example:

MATCH (n)-->(b) RETURN b

The variables are n and b.

Variable names are case sensitive, and can contain underscores and alphanumeric characters (a-z, 0-9), but must always start with a letter. If other characters are needed, you can quote the variable usingbackquote (`) signs.

The same rules apply to property names.

Variables are only visible in the same query part

Variables are not carried over to subsequent queries. If multiple query parts arechained together using WITH, variables have to be listed in the WITH clause to becarried over to the next part. For more information see With.

6.4. Operators

6.4.1. Mathematical operators

The mathematical operators are +, -, *, / and %, ^.

6.4.2. Comparison operators

The comparison operators are =, <>, <, >, ⇐, >=, IS NULL, and IS NOT NULL. See Equality and Comparisonof Values on how they behave.

The operators STARTS WITH, ENDS WITH and CONTAINS can be used to search for a string value by itscontent.

6.4.3. Boolean operators

The boolean operators are AND, OR, XOR, NOT.

37

6.4.4. String operators

Strings can be concatenated using the + operator. For regular expression matching the =~ operator isused.

6.4.5. List operators

Lists can be concatenated using the + operator. To check if an element exists in a list, you can use theIN operator.

6.4.6. Property operators

Since version 2.0, the previously existing property operators ? and ! have beenremoved. This syntax is no longer supported. Missing properties are now returnedas NULL. Please use (NOT(has(<ident>.prop)) OR <ident>.prop=<value>) if youreally need the old behavior of the ? operator. — Also, the use of ? for optionalrelationships has been removed in favor of the newly introduced OPTIONAL MATCHclause.

6.4.7. Equality and Comparison of Values

Equality

Cypher supports comparing values (see Values) by equality using the = and <> operators.

Values of the same type are only equal if they are the same identical value (e.g. 3 = 3 and "x" <>"xy").

Maps are only equal if they map exactly the same keys to equal values and lists are only equal if theycontain the same sequence of equal values (e.g. [3, 4] = [1+2, 8/2]).

Values of different types are considered as equal according to the following rules:

• Paths are treated as lists of alternating nodes and relationships and are equal to all lists thatcontain that very same sequence of nodes and relationships.

• Testing any value against NULL with both the = and the <> operators always is NULL. This includesNULL = NULL and NULL <> NULL. The only way to reliably test if a value v is NULL is by using thespecial v IS NULL, or v IS NOT NULL equality operators.

All other combinations of types of values cannot be compared with each other. Especially, nodes,relationships, and literal maps are incomparable with each other.

It is an error to compare values that cannot be compared.

6.4.8. Ordering and Comparison of Values

The comparison operators ⇐, < (for ascending) and >=, > (for descending) are used to compare valuesfor ordering. The following points give some details on how the comparison is performed.

• Numerical values are compared for ordering using numerical order (e.g. 3 < 4 is true).

• The special value java.lang.Double.NaN is regarded as being larger than all other numbers.

• String values are compared for ordering using lexicographic order (e.g. "x" < "xy").

• Boolean values are compared for ordering such that false < true.

38

• Comparing for ordering when one argument is NULL is NULL (e.g. NULL < 3 is NULL).

• It is an error to compare other types of values with each other for ordering.

6.4.9. Chaining Comparison Operations

Comparisons can be chained arbitrarily, e.g., x < y ⇐ z is equivalent to x < y AND y ⇐ z.

Formally, if a, b, c, …, y, z are expressions and op1, op2, …, opN are comparison operators, then aop1 b op2 c … y opN z is equivalent to a op1 b and b op2 c and … y opN z.

Note that a op1 b op2 c does not imply any kind of comparison between a and c, so that, e.g., x < y >z is perfectly legal (though perhaps not pretty).

The example:

MATCH (n) WHERE 21 < n.age <= 30 RETURN n

is equivalent to

MATCH (n) WHERE 21 < n.age AND n.age <= 30 RETURN n

Thus it will match all nodes where the age is between 21 and 30.

This syntax extends to all equality and inequality comparisons, as well as extending to chains longerthan three.

For example:

a < b = c <= d <> e

Is equivalent to:

a < b AND b = c AND c <= d AND d <> e

For other comparison operators, see Comparison operators.

6.5. Comments

To add comments to your queries, use double slash. Examples:

MATCH (n) RETURN n //This is an end of line comment

MATCH (n)//This is a whole line commentRETURN n

MATCH (n) WHERE n.property = "//This is NOT a comment" RETURN n

39

6.6. Patterns

Patterns and pattern-matching are at the very heart of Cypher, so being effective with Cypher requiresa good understanding of patterns.

Using patterns, you describe the shape of the data you’re looking for. For example, in the MATCH clauseyou describe the shape with a pattern, and Cypher will figure out how to get that data for you.

The pattern describes the data using a form that is very similar to how one typically draws the shapeof property graph data on a whiteboard: usually as circles (representing nodes) and arrows betweenthem to represent relationships.

Patterns appear in multiple places in Cypher: in MATCH, CREATE and MERGE clauses, and in patternexpressions. Each of these is described in more details in:

• Match

• Optional Match

• Create

• Merge

• Using path patterns in WHERE

6.6.1. Patterns for nodes

The very simplest ``shape'' that can be described in a pattern is a node. A node is described using apair of parentheses, and is typically given a name. For example:

(a)

This simple pattern describes a single node, and names that node using the variable a.

6.6.2. Patterns for related nodes

More interesting is patterns that describe multiple nodes and relationships between them. Cypherpatterns describe relationships by employing an arrow between two nodes. For example:

(a)-->(b)

This pattern describes a very simple data shape: two nodes, and a single relationship from one to theother. In this example, the two nodes are both named as a and b respectively, and the relationship is`directed'': it goes from `a to b.

This way of describing nodes and relationships can be extended to cover an arbitrary number ofnodes and the relationships between them, for example:

(a)-->(b)<--(c)

Such a series of connected nodes and relationships is called a "path".

Note that the naming of the nodes in these patterns is only necessary should one need to refer to thesame node again, either later in the pattern or elsewhere in the Cypher query. If this is not necessarythen the name may be omitted, like so:

40

(a)-->()<--(c)

6.6.3. Labels

In addition to simply describing the shape of a node in the pattern, one can also describe attributes.The most simple attribute that can be described in the pattern is a label that the node must have. Forexample:

(a:User)-->(b)

One can also describe a node that has multiple labels:

(a:User:Admin)-->(b)

6.6.4. Specifying properties

Nodes and relationships are the fundamental structures in a graph. Neo4j uses properties on both ofthese to allow for far richer models.

Properties can be expressed in patterns using a map-construct: curly brackets surrounding a numberof key-expression pairs, separated by commas. E.g. a node with two properties on it would look like:

(a { name: "Andres", sport: "Brazilian Ju-Jitsu" })

A relationship with expectations on it would could look like:

(a)-[{blocked: false}]->(b)

When properties appear in patterns, they add an additional constraint to the shape of the data. In thecase of a CREATE clause, the properties will be set in the newly created nodes and relationships. In thecase of a MERGE clause, the properties will be used as additional constraints on the shape any existingdata must have (the specified properties must exactly match any existing data in the graph). If nomatching data is found, then MERGE behaves like CREATE and the properties will be set in the newlycreated nodes and relationships.

Note that patterns supplied to CREATE may use a single parameter to specify properties, e.g: CREATE(node {paramName}). This is not possible with patterns used in other clauses, as Cypher needs to knowthe property names at the time the query is compiled, so that matching can be done effectively.

6.6.5. Describing relationships

The simplest way to describe a relationship is by using the arrow between two nodes, as in theprevious examples. Using this technique, you can describe that the relationship should exist and thedirectionality of it. If you don’t care about the direction of the relationship, the arrow head can beomitted, like so:

(a)--(b)

As with nodes, relationships may also be given names. In this case, a pair of square brackets is used tobreak up the arrow and the variable is placed between. For example:

41

(a)-[r]->(b)

Much like labels on nodes, relationships can have types. To describe a relationship with a specific type,you can specify this like so:

(a)-[r:REL_TYPE]->(b)

Unlike labels, relationships can only have one type. But if we’d like to describe some data such that therelationship could have any one of a set of types, then they can all be listed in the pattern, separatingthem with the pipe symbol | like this:

(a)-[r:TYPE1|TYPE2]->(b)

Note that this form of pattern can only be used to describe existing data (ie. when using a pattern withMATCH or as an expression). It will not work with CREATE or MERGE, since it’s not possible to create arelationship with multiple types.

As with nodes, the name of the relationship can always be omitted, in this case like so:

(a)-[:REL_TYPE]->(b)

Variable length

Variable length pattern matching in versions 2.1.x and earlier does not enforcerelationship uniqueness for patterns described inside of a single MATCH clause. Thismeans that a query such as the following: MATCH (a)-[r]→(b), (a)-[rs*]→(c)RETURN may include r as part of the rs set. This behavior has changed inversions 2.2.0 and later, in such a way that r will be excluded from theresult set, as this better adheres to the rules of relationship uniqueness asdocumented here Uniqueness. If you have a query pattern that needs to retracerelationships rather than ignoring them as the relationship uniqueness rulesnormally dictate, you can accomplish this using multiple match clauses, asfollows: MATCH (a)-[r]→(b) MATCH (a)-[rs]→(c) RETURN *. This will work in allversions of Neo4j that support the MATCH clause, namely 2.0.0 and later.

Rather than describing a long path using a sequence of many node and relationship descriptions in apattern, many relationships (and the intermediate nodes) can be described by specifying a length inthe relationship description of a pattern. For example:

(a)-[*2]->(b)

This describes a graph of three nodes and two relationship, all in one path (a path of length 2). This isequivalent to:

(a)-->()-->(b)

A range of lengths can also be specified: such relationship patterns are called ``variable lengthrelationships''. For example:

(a)-[*3..5]->(b)

42

This is a minimum length of 3, and a maximum of 5. It describes a graph of either 4 nodes and 3relationships, 5 nodes and 4 relationships or 6 nodes and 5 relationships, all connected together in asingle path.

Either bound can be omitted. For example, to describe paths of length 3 or more, use:

(a)-[*3..]->(b)

And to describe paths of length 5 or less, use:

(a)-[*..5]->(b)

Both bounds can be omitted, allowing paths of any length to be described:

(a)-[*]->(b)

As a simple example, let’s take the query below:

Query

MATCH (me)-[:KNOWS*1..2]-(remote_friend)WHERE me.name = "Filipa"RETURN remote_friend.name

Result

+--------------------+| remote_friend.name |+--------------------+| "Dilshad" || "Anders" |+--------------------+2 rows

This query finds data in the graph which a shape that fits the pattern: specifically a node (with thename property Filipa) and then the KNOWS related nodes, one or two steps out. This is a typicalexample of finding first and second degree friends.

Note that variable length relationships can not be used with CREATE and MERGE.

6.6.6. Assigning to path variables

As described above, a series of connected nodes and relationships is called a "path". Cypher allowspaths to be named using an identifer, like so:

p = (a)-[*3..5]->(b)

You can do this in MATCH, CREATE and MERGE, but not when using patterns as expressions.

6.7. Lists

Cypher has good support for lists.

43

6.7.1. Lists in general

A literal list is created by using brackets and separating the elements in the list with commas.

Query

RETURN [0,1,2,3,4,5,6,7,8,9] AS list

Result

+-----------------------+| list |+-----------------------+| [0,1,2,3,4,5,6,7,8,9] |+-----------------------+1 row

In our examples, we’ll use the range function. It gives you a list containing all numbers between givenstart and end numbers. Range is inclusive in both ends.

To access individual elements in the list, we use the square brackets again. This will extract from thestart index and up to but not including the end index.

Query

RETURN range(0,10)[3]

Result

+----------------+| range(0,10)[3] |+----------------+| 3 |+----------------+1 row

You can also use negative numbers, to start from the end of the list instead.

Query

RETURN range(0,10)[-3]

Result

+-----------------+| range(0,10)[-3] |+-----------------+| 8 |+-----------------+1 row

Finally, you can use ranges inside the brackets to return ranges of the list.

Query

RETURN range(0,10)[0..3]

44

Result

+-------------------+| range(0,10)[0..3] |+-------------------+| [0,1,2] |+-------------------+1 row

Query

RETURN range(0,10)[0..-5]

Result

+--------------------+| range(0,10)[0..-5] |+--------------------+| [0,1,2,3,4,5] |+--------------------+1 row

Query

RETURN range(0,10)[-5..]

Result

+-------------------+| range(0,10)[-5..] |+-------------------+| [6,7,8,9,10] |+-------------------+1 row

Query

RETURN range(0,10)[..4]

Result

+------------------+| range(0,10)[..4] |+------------------+| [0,1,2,3] |+------------------+1 row

Out-of-bound slices are simply truncated, but out-of-bound single elements returnNULL.

Query

RETURN range(0,10)[15]

45

Result

+-----------------+| range(0,10)[15] |+-----------------+| <null> |+-----------------+1 row

Query

RETURN range(0,10)[5..15]

Result

+--------------------+| range(0,10)[5..15] |+--------------------+| [5,6,7,8,9,10] |+--------------------+1 row

You can get the size of a list like this:

Query

RETURN size(range(0,10)[0..3])

Result

+-------------------------+| size(range(0,10)[0..3]) |+-------------------------+| 3 |+-------------------------+1 row

6.7.2. List comprehension

List comprehension is a syntactic construct available in Cypher for creating a list based on existinglists. It follows the form of the mathematical set-builder notation (set comprehension) instead of theuse of map and filter functions.

Query

RETURN [x IN range(0,10) WHERE x % 2 = 0 | x^3] AS result

Result

+-----------------------------------+| result |+-----------------------------------+| [0.0,8.0,64.0,216.0,512.0,1000.0] |+-----------------------------------+1 row

Either the WHERE part, or the expression, can be omitted, if you only want to filter or maprespectively.

46

Query

RETURN [x IN range(0,10) WHERE x % 2 = 0] AS result

Result

+----------------+| result |+----------------+| [0,2,4,6,8,10] |+----------------+1 row

Query

RETURN [x IN range(0,10)| x^3] AS result

Result

+--------------------------------------------------------------+| result |+--------------------------------------------------------------+| [0.0,1.0,8.0,27.0,64.0,125.0,216.0,343.0,512.0,729.0,1000.0] |+--------------------------------------------------------------+1 row

6.7.3. Literal maps

From Cypher, you can also construct maps. Through REST you will get JSON objects; in Java they will bejava.util.Map<String,Object>.

Query

RETURN { key : "Value", listKey: [{ inner: "Map1" }, { inner: "Map2" }]} AS result

Result

+--------------------------------------------------------------------+| result |+--------------------------------------------------------------------+| {key -> "Value", listKey -> [{inner -> "Map1"},{inner -> "Map2"}]} |+--------------------------------------------------------------------+1 row

6.8. Working with NULL

6.8.1. Introduction to NULL in Cypher

In Cypher, NULL is used to represent missing or undefined values. Conceptually, NULL means ``amissing unknown value'' and it is treated somewhat differently from other values. For example gettinga property from a node that does not have said property produces NULL. Most expressions that takeNULL as input will produce NULL. This includes boolean expressions that are used as predicates in theWHERE clause. In this case, anything that is not TRUE is interpreted as being false.

NULL is not equal to NULL. Not knowing two values does not imply that they are the same value. Sothe expression NULL = NULL yields NULL and not TRUE.

47

6.8.2. Logical operations with NULL

The logical operators (AND, OR, XOR, IN, NOT) treat NULL as the ``unknown'' value of three-valuedlogic. Here is the truth table for AND, OR and XOR.

a b a AND b a OR b a XOR b

FALSE FALSE FALSE FALSE FALSE

FALSE NULL FALSE NULL NULL

FALSE TRUE FALSE TRUE TRUE

TRUE FALSE FALSE TRUE TRUE

TRUE NULL NULL TRUE NULL

TRUE TRUE TRUE TRUE FALSE

NULL FALSE FALSE NULL NULL

NULL NULL NULL NULL NULL

NULL TRUE NULL TRUE NULL

6.8.3. The IN operator and NULL

The IN operator follows similar logic. If Cypher knows that something exists in a list, the result will betrue. Any list that contains a NULL and doesn’t have a matching element will return NULL. Otherwise,the result will be false. Here is a table with examples:

Expression Result

2 IN [1, 2, 3] TRUE

2 IN [1, NULL, 3] NULL

2 IN [1, 2, NULL] TRUE

2 IN [1] FALSE

2 IN [] FALSE

NULL IN [1,2,3] NULL

NULL IN [1,NULL,3] NULL

NULL IN [] FALSE

Using ALL, ANY, NONE, and SINGLE follows a similar rule. If the result can be calculated definitely,TRUE or FALSE is returned. Otherwise NULL is produced.

6.8.4. Expressions that return NULL

• Getting a missing element from a list: [][0], head([])

• Trying to access a property that does not exist on a node or relationship: n.missingProperty

• Comparisons when either side is NULL: `1 < NULL`

• Arithmetic expressions containing NULL: `1 + NULL`

• Function calls where any arguments are NULL: sin(NULL)

48

Chapter 7. General Clauses

7.1. Return

The RETURN clause defines what to include in the query result set.

In the RETURN part of your query, you define which parts of the pattern you are interested in. It can benodes, relationships, or properties on these.

If what you actually want is the value of a property, make sure to not return the fullnode/relationship. This will improve performance.

name = 'A'happy = 'Yes!'age = 55

name = 'B'

BLOCKS KNOWS

Figure 2. Graph

7.1.1. Return nodes

To return a node, list it in the RETURN statement.

Query

MATCH (n { name: "B" })RETURN n

The example will return the node.

Result

+-------------------+| n |+-------------------+| Node[1]{name:"B"} |+-------------------+1 row

7.1.2. Return relationships

To return a relationship, just include it in the RETURN list.

Query

MATCH (n { name: "A" })-[r:KNOWS]->(c)RETURN r

The relationship is returned by the example.

49

Result

+-------------+| r |+-------------+| :KNOWS[0]{} |+-------------+1 row

7.1.3. Return property

To return a property, use the dot separator, like this:

Query

MATCH (n { name: "A" })RETURN n.name

The value of the property name gets returned.

Result

+--------+| n.name |+--------+| "A" |+--------+1 row

7.1.4. Return all elements

When you want to return all nodes, relationships and paths found in a query, you can use the *symbol.

Query

MATCH p=(a { name: "A" })-[r]->(b)RETURN *

This returns the two nodes, the relationship and the path used in the query.

Result

+---------------------------------------------------------------------------------------------------------------------------------------------------+| a | b | p| r |+---------------------------------------------------------------------------------------------------------------------------------------------------+| Node[0]{name:"A",happy:"Yes!",age:55} | Node[1]{name:"B"} |[Node[0]{name:"A",happy:"Yes!",age:55},:BLOCKS[1]{},Node[1]{name:"B"}] | :BLOCKS[1]{} || Node[0]{name:"A",happy:"Yes!",age:55} | Node[1]{name:"B"} |[Node[0]{name:"A",happy:"Yes!",age:55},:KNOWS[0]{},Node[1]{name:"B"}] | :KNOWS[0]{} |+---------------------------------------------------------------------------------------------------------------------------------------------------+2 rows

7.1.5. Variable with uncommon characters

To introduce a placeholder that is made up of characters that are outside of the english alphabet,you can use the ` to enclose the variable, like this:

50

Query

MATCH (`This isn't a common variable`)WHERE `This isn't a common variable`.name='A'RETURN `This isn't a common variable`.happy

The node with name "A" is returned

Result

+--------------------------------------+| `This isn't a common variable`.happy |+--------------------------------------+| "Yes!" |+--------------------------------------+1 row

7.1.6. Column alias

If the name of the column should be different from the expression used, you can rename it by usingAS <new name>.

Query

MATCH (a { name: "A" })RETURN a.age AS SomethingTotallyDifferent

Returns the age property of a node, but renames the column.

Result

+---------------------------+| SomethingTotallyDifferent |+---------------------------+| 55 |+---------------------------+1 row

7.1.7. Optional properties

If a property might or might not be there, you can still select it as usual. It will be treated as NULL if itis missing

Query

MATCH (n)RETURN n.age

This example returns the age when the node has that property, or null if the property is not there.

Result

+--------+| n.age |+--------+| 55 || <null> |+--------+2 rows

51

7.1.8. Other expressions

Any expression can be used as a return item — literals, predicates, properties, functions, andeverything else.

Query

MATCH (a { name: "A" })RETURN a.age > 30, "I'm a literal",(a)-->()

Returns a predicate, a literal and function call with a pattern expression parameter.

Result

+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+| a.age > 30 | "I'm a literal" | (a)-->()|+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+| true | "I'm a literal" |[[Node[0]{name:"A",happy:"Yes!",age:55},:BLOCKS[1]{},Node[1]{name:"B"}],[Node[0]{name:"A",happy:"Yes!",age:55},:KNOWS[0]{},Node[1]{name:"B"}]] |+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+1 row

7.1.9. Unique results

DISTINCT retrieves only unique rows depending on the columns that have been selected to output.

Query

MATCH (a { name: "A" })-->(b)RETURN DISTINCT b

The node named B is returned by the query, but only once.

Result

+-------------------+| b |+-------------------+| Node[1]{name:"B"} |+-------------------+1 row

7.2. Order by

ORDER BY is a sub-clause following RETURN or WITH, and it specifies that the output should be

sorted and how.

Note that you can not sort on nodes or relationships, just on properties on these. ORDER BY relies oncomparisons to sort the output, see Ordering and Comparison of Values.

In terms of scope of variables, ORDER BY follows special rules, depending on if the projecting RETURN orWITH clause is either aggregating or DISTINCT. If it is an aggregating or DISTINCT projection, only thevariables available in the projection are available. If the projection does not alter the output cardinality(which aggregation and DISTINCT do), variables available from before the projecting clause are also

52

available. When the projection clause shadows already existing variables, only the new variables areavailable.

Lastly, it is not allowed to use aggregating expressions in the ORDER BY sub-clause if they are not alsolisted in the projecting clause. This last rule is to make sure that ORDER BY does not change the results,only the order of them.

name = 'A'age = 34length = 170

name = 'B'age = 34

KNOWS

name = 'C'age = 32length = 185

KNOWS

Figure 3. Graph

7.2.1. Order nodes by property

ORDER BY is used to sort the output.

Query

MATCH (n)RETURN nORDER BY n.name

The nodes are returned, sorted by their name.

Result

+-------------------------------------+| n |+-------------------------------------+| Node[0]{name:"A",age:34,length:170} || Node[1]{name:"B",age:34} || Node[2]{name:"C",age:32,length:185} |+-------------------------------------+3 rows

7.2.2. Order nodes by multiple properties

You can order by multiple properties by stating each variable in the ORDER BY clause. Cypher will sortthe result by the first variable listed, and for equals values, go to the next property in the ORDER BYclause, and so on.

Query

MATCH (n)RETURN nORDER BY n.age, n.name

This returns the nodes, sorted first by their age, and then by their name.

53

Result

+-------------------------------------+| n |+-------------------------------------+| Node[2]{name:"C",age:32,length:185} || Node[0]{name:"A",age:34,length:170} || Node[1]{name:"B",age:34} |+-------------------------------------+3 rows

7.2.3. Order nodes in descending order

By adding DESC[ENDING] after the variable to sort on, the sort will be done in reverse order.

Query

MATCH (n)RETURN nORDER BY n.name DESC

The example returns the nodes, sorted by their name reversely.

Result

+-------------------------------------+| n |+-------------------------------------+| Node[2]{name:"C",age:32,length:185} || Node[1]{name:"B",age:34} || Node[0]{name:"A",age:34,length:170} |+-------------------------------------+3 rows

7.2.4. Ordering NULL

When sorting the result set, NULL will always come at the end of the result set for ascending sorting,and first when doing descending sort.

Query

MATCH (n)RETURN n.length, nORDER BY n.length

The nodes are returned sorted by the length property, with a node without that property last.

Result

+------------------------------------------------+| n.length | n |+------------------------------------------------+| 170 | Node[0]{name:"A",age:34,length:170} || 185 | Node[2]{name:"C",age:32,length:185} || <null> | Node[1]{name:"B",age:34} |+------------------------------------------------+3 rows

7.3. Limit

LIMIT constrains the number of rows in the output.

54

LIMIT accepts any expression that evaluates to a positive integer — however the expression cannotrefer to nodes or relationships.

name = 'A'

name = 'E'

KNOWS

name = 'D'

KNOWS

name = 'C'

KNOWS

name = 'B'

KNOWS

Figure 4. Graph

7.3.1. Return first part

To return a subset of the result, starting from the top, use this syntax:

Query

MATCH (n)RETURN nORDER BY n.nameLIMIT 3

The top three items are returned by the example query.

Result

+-------------------+| n |+-------------------+| Node[0]{name:"A"} || Node[1]{name:"B"} || Node[2]{name:"C"} |+-------------------+3 rows

7.3.2. Return first from expression

Limit accepts any expression that evaluates to a positive integer as long as it is not referring to anyexternal variables:

Parameters

{ "p" : 12}

Query

MATCH (n)RETURN nORDER BY n.nameLIMIT toInt(3 * rand())+ 1

Returns one to three top items

55

Result

+-------------------+| n |+-------------------+| Node[0]{name:"A"} || Node[1]{name:"B"} |+-------------------+2 rows

7.4. Skip

SKIP defines from which row to start including the rows in the output.

By using SKIP, the result set will get trimmed from the top. Please note that no guarantees are madeon the order of the result unless the query specifies the ORDER BY clause. SKIP accepts any expressionthat evaluates to a positive integer — however the expression cannot refer to nodes or relationships.

name = 'A'

name = 'E'

KNOWS

name = 'D'

KNOWS

name = 'C'

KNOWS

name = 'B'

KNOWS

Figure 5. Graph

7.4.1. Skip first three

To return a subset of the result, starting from the fourth result, use the following syntax:

Query

MATCH (n)RETURN nORDER BY n.nameSKIP 3

The first three nodes are skipped, and only the last two are returned in the result.

Result

+-------------------+| n |+-------------------+| Node[3]{name:"D"} || Node[4]{name:"E"} |+-------------------+2 rows

7.4.2. Return middle two

To return a subset of the result, starting from somewhere in the middle, use this syntax:

56

Query

MATCH (n)RETURN nORDER BY n.nameSKIP 1LIMIT 2

Two nodes from the middle are returned.

Result

+-------------------+| n |+-------------------+| Node[1]{name:"B"} || Node[2]{name:"C"} |+-------------------+2 rows

7.4.3. Skip first from expression

Skip accepts any expression that evaluates to a positive integer as long as it is not referring to anyexternal variables:

Query

MATCH (n)RETURN nORDER BY n.nameSKIP toInt(3*rand())+ 1

The first three nodes are skipped, and only the last two are returned in the result.

Result

+-------------------+| n |+-------------------+| Node[2]{name:"C"} || Node[3]{name:"D"} || Node[4]{name:"E"} |+-------------------+3 rows

7.5. With

The WITH clause allows query parts to be chained together, piping the results from one to be

used as starting points or criteria in the next.

Using WITH, you can manipulate the output before it is passed on to the following query parts. Themanipulations can be of the shape and/or number of entries in the result set.

One common usage of WITH is to limit the number of entries that are then passed on to other MATCHclauses. By combining ORDER BY and LIMIT, it’s possible to get the top X entries by some criteria, andthen bring in additional data from the graph.

Another use is to filter on aggregated values. WITH is used to introduce aggregates which can then byused in predicates in WHERE. These aggregate expressions create new bindings in the results. WITH canalso, like RETURN, alias expressions that are introduced into the results using the aliases as binding

57

name.

WITH is also used to separate reading from updating of the graph. Every part of a query must be eitherread-only or write-only. When going from a writing part to a reading part, the switch must be donewith a WITH clause.

name = 'Anders'

name = 'Ceasar'

BLOCKS

name = 'Bossman'

KNOWS

name = 'Emil'

KNOWS

name = 'David'

BLOCKSKNOWS

KNOWS

Figure 6. Graph

7.5.1. Filter on aggregate function results

Aggregated results have to pass through a WITH clause to be able to filter on.

Query

MATCH (david { name: "David" })--(otherPerson)-->()WITH otherPerson, count(*) AS foafWHERE foaf > 1RETURN otherPerson

The person connected to David with the at least more than one outgoing relationship will be returnedby the query.

Result

+------------------------+| otherPerson |+------------------------+| Node[0]{name:"Anders"} |+------------------------+1 row

7.5.2. Sort results before using collect on them

You can sort your results before passing them to collect, thus sorting the resulting list.

Query

MATCH (n)WITH nORDER BY n.name DESC LIMIT 3RETURN collect(n.name)

A list of the names of people in reverse order, limited to 3, in a list.

58

Result

+---------------------------+| collect(n.name) |+---------------------------+| ["Emil","David","Ceasar"] |+---------------------------+1 row

7.5.3. Limit branching of your path search

You can match paths, limit to a certain number, and then match again using those paths as a base Aswell as any number of similar limited searches.

Query

MATCH (n { name: "Anders" })--(m)WITH mORDER BY m.name DESC LIMIT 1MATCH (m)--(o)RETURN o.name

Starting at Anders, find all matching nodes, order by name descending and get the top result, thenfind all the nodes connected to that top result, and return their names.

Result

+-----------+| o.name |+-----------+| "Bossman" || "Anders" |+-----------+2 rows

7.6. Unwind

UNWIND expands a list into a sequence of rows.

With UNWIND, you can transform any list back into individual rows. These lists can be parameters thatwere passed in, previously COLLECTed result or other list expressions.

One common usage of unwind is to create distinct lists. Another is to create data from parameter liststhat are provided to the query.

UNWIND requires you to specify a new name for the inner values.

7.6.1. Unwind a list

We want to transform the literal list into rows named x and return them.

Query

UNWIND[1,2,3] AS xRETURN x

Each value of the original list is returned as an individual row.

59

Result

+---+| x |+---+| 1 || 2 || 3 |+---+3 rows

7.6.2. Create a distinct list

We want to transform a list of duplicates into a set using DISTINCT.

Query

WITH [1,1,2,2] AS coll UNWIND coll AS xWITH DISTINCT xRETURN collect(x) AS SET

Each value of the original list is unwound and passed through DISTINCT to create a unique set.

Result

+-------+| set |+-------+| [1,2] |+-------+1 row

7.6.3. Create nodes from a list parameter

Create a number of nodes and relationships from a parameter-list without using FOREACH.

Parameters

{ "events" : [ { "year" : 2014, "id" : 1 }, { "year" : 2014, "id" : 2 } ]}

Query

UNWIND { events } AS eventMERGE (y:Year { year:event.year })MERGE (y)<-[:IN]-(e:Event { id:event.id })RETURN e.id AS xORDER BY x

Each value of the original list is unwound and passed through MERGE to find or create the nodes andrelationships.

60

Result

+---+| x |+---+| 1 || 2 |+---+2 rowsNodes created: 3Relationships created: 2Properties set: 3Labels added: 3

7.7. Union

The UNION clause is used to combine the result of multiple queries.

It combines the results of two or more queries into a single result set that includes all the rows thatbelong to all queries in the union.

The number and the names of the columns must be identical in all queries combined by using UNION.

To keep all the result rows, use UNION ALL. Using just UNION will combine and remove duplicates fromthe result set.

Actor

name = 'Anthony Hopkins'

Movie

title = 'Hitchcock'

ACTS_INActor

name = 'Helen Mirren'

KNOWS

ACTS_IN

Actor

name = 'Hitchcock'

Figure 7. Graph

7.7.1. Combine two queries

Combining the results from two queries is done using UNION ALL.

Query

MATCH (n:Actor)RETURN n.name AS nameUNION ALL MATCH (n:Movie)RETURN n.title AS name

The combined result is returned, including duplicates.

61

Result

+-------------------+| name |+-------------------+| "Anthony Hopkins" || "Helen Mirren" || "Hitchcock" || "Hitchcock" |+-------------------+4 rows

7.7.2. Combine two queries and remove duplicates

By not including ALL in the UNION, duplicates are removed from the combined result set

Query

MATCH (n:Actor)RETURN n.name AS nameUNIONMATCH (n:Movie)RETURN n.title AS name

The combined result is returned, without duplicates.

Result

+-------------------+| name |+-------------------+| "Anthony Hopkins" || "Helen Mirren" || "Hitchcock" |+-------------------+3 rows

7.8. Call

The CALL clause is used to call a procedure deployed in the database.

The examples showing how to use arguments when invoking procedures all use the followingprocedure:

62

public class IndexingProcedure{ @Context public GraphDatabaseService db;

/** * Adds a node to a named legacy index. Useful to, for instance, update * a full-text index through cypher. * @param indexName the name of the index in question * @param nodeId id of the node to add to the index * @param propKey property to index (value is read from the node) */ @Procedure @PerformsWrites public void addNodeToIndex( @Name("indexName") String indexName, @Name("node") long nodeId, @Name("propKey" ) String propKey ) { Node node = db.getNodeById( nodeId ); db.index() .forNodes( indexName ) .add( node, propKey, node.getProperty( propKey ) ); }}

This clause cannot be combined with other clauses.

7.8.1. Call a procedure

This invokes the built-in procedure 'db.labels', which lists all in-use labels in the database.

Query

CALL db.labels

Result

+-----------------+| label |+-----------------+| "User" || "Administrator" |+-----------------+2 rows

7.8.2. Call a procedure with literal arguments

This invokes the example procedure org.neo4j.procedure.example.addNodeToIndex using argumentsthat are written out directly in the statement text. This is called literal arguments.

Query

CALL org.neo4j.procedure.example.addNodeToIndex('users', 0, 'name')

Since our example procedure does not return any result, the result is empty.

Result

+--------------------------------------------+| No data returned, and nothing was changed. |+--------------------------------------------+

63

7.8.3. Call a procedure with parameter arguments

This invokes the example procedure org.neo4j.procedure.example.addNodeToIndex using parameters.The procedure arguments are satisfied by matching the parameter keys to the named procedurearguments.

Parameters

{ "indexName" : "users", "node" : 0, "propKey" : "name"}

Query

CALL org.neo4j.procedure.example.addNodeToIndex


Result


7.8.4. Call a procedure with mixed literal and parameter arguments

This invokes the example procedure org.neo4j.procedure.example.addNodeToIndex using both literaland parameterized arguments.

Parameters

{ "node" : 0}

Query

CALL org.neo4j.procedure.example.addNodeToIndex('users', { node }, 'name')


Result


7.8.5. Call a procedure within a complex query

This invokes the built-in procedure 'db.labels' to count all in-use labels in the database

Query

CALL db.labels() YIELD labelRETURN count(label) AS numLabels

64

Since the procedure call is part of a larger query, all outputs must be named explicitly

Result

+-----------+| numLabels |+-----------+| 2 |+-----------+1 row

7.8.6. Call a procedure within a complex query and rename itsoutputs

This invokes the built-in procedure 'db.propertyKeys' as part of counting the number of nodes perproperty key in-use in the database

Query

CALL db.propertyKeys() YIELD propertyKey AS propMATCH (n)WHERE n[prop] IS NOT NULL RETURN prop, count(n) AS numNodes

Since the procedure call is part of a larger query, all outputs must be named explicitly

Result

+-------------------+| prop | numNodes |+-------------------+| "name" | 1 |+-------------------+1 row

65

Chapter 8. Reading ClausesThe flow of data within a Cypher query is an unordered sequence of maps with key-value pairs — a setof possible bindings between the variables in the query and values derived from the database. Thisset is refined and augmented by subsequent parts of the query.

8.1. Match

The MATCH clause is used to search for the pattern described in it.

8.1.1. Introduction

The MATCH clause allows you to specify the patterns Neo4j will search for in the database. This is theprimary way of getting data into the current set of bindings. It is worth reading up more on thespecification of the patterns themselves in Patterns.

MATCH is often coupled to a WHERE part which adds restrictions, or predicates, to the MATCH patterns,making them more specific. The predicates are part of the pattern description, and should not beconsidered a filter applied only after the matching is done. This means that WHERE should always be puttogether with the MATCH clause it belongs to.

MATCH can occur at the beginning of the query or later, possibly after a WITH. If it is the first clause,nothing will have been bound yet, and Neo4j will design a search to find the results matching theclause and any associated predicates specified in any WHERE part. This could involve a scan of thedatabase, a search for nodes of a certain label, or a search of an index to find starting points for thepattern matching. Nodes and relationships found by this search are available as bound patternelements, and can be used for pattern matching of sub-graphs. They can also be used in any furtherMATCH clauses, where Neo4j will use the known elements, and from there find further unknownelements.

Cypher is declarative, and so usually the query itself does not specify the algorithm to use to performthe search. Neo4j will automatically work out the best approach to finding start nodes and matchingpatterns. Predicates in WHERE parts can be evaluated before pattern matching, during patternmatching, or after finding matches. However, there are cases where you can influence the decisionstaken by the query compiler. Read more about indexes in Indexes, and more about specifying hints toforce Neo4j to solve a query in a specific way in Using.

To understand more about the patterns used in the MATCH clause, read Patterns

The following graph is used for the examples below:

Person

name = 'Charlie Sheen'

Movie

title = 'Wall Street'

ACTED_INrole = 'Bud Fox'

Person

name = 'Martin Sheen'

ACTED_INrole = 'Carl Fox'

Movie

title = 'The American President'

ACTED_INrole = 'A.J. MacInerney'

Person

name = 'Michael Douglas'

ACTED_INrole = 'Gordon Gekko'

ACTED_INrole = 'President Andrew Shepherd'

Person

name = 'Oliver Stone'

DIRECTED

Person

name = 'Rob Reiner'

DIRECTED

Figure 8. Graph

8.1.2. Basic node finding

Get all nodes

By just specifying a pattern with a single node and no labels, all nodes in the graph will be returned.

66

Query

MATCH (n)RETURN n

Returns all the nodes in the database.

Table 2. Result

n

Node[0]\{name:"Charlie Sheen"\}

Node[1]\{name:"Martin Sheen"\}

Node[2]\{name:"Michael Douglas"\}

Node[3]\{name:"Oliver Stone"\}

Node[4]\{name:"Rob Reiner"\}

Node[5]\{title:"Wall Street"\}

Node[6]\{title:"The American President"\}

7 rows

Get all nodes with a label

Getting all nodes with a label on them is done with a single node pattern where the node has a labelon it.

Query

MATCH (movie:Movie)RETURN movie.title

Returns all the movies in the database.

Table 3. Result

movie.title

"Wall Street"

"The American President"

2 rows

Related nodes

The symbol -- means related to, without regard to type or direction of the relationship.

Query

MATCH (director { name:'Oliver Stone' })--(movie)RETURN movie.title

Returns all the movies directed by 'Oliver Stone'.

Table 4. Result

movie.title

"Wall Street"

67

movie.title

1 row

Match with labels

To constrain your pattern with labels on nodes, you add it to your pattern nodes, using the labelsyntax.

Query

MATCH (:Person { name:'Oliver Stone' })--(movie:Movie)RETURN movie.title

Returns any nodes connected with the Person Oliver that are labeled Movie.

Table 5. Result

movie.title

"Wall Street"

1 row

8.1.3. Relationship basics

Outgoing relationships

When the direction of a relationship is interesting, it is shown by using -→ or ←-, like this:

Query

MATCH (:Person { name:'Oliver Stone' })-->(movie)RETURN movie.title

Returns any nodes connected with the Person Oliver by an outgoing relationship.

Table 6. Result

movie.title

"Wall Street"

1 row

Directed relationships and variable

If an variable is needed, either for filtering on properties of the relationship, or to return therelationship, this is how you introduce the variable.

Query

MATCH (:Person { name:'Oliver Stone' })-[r]->(movie)RETURN type(r)

Returns the type of each outgoing relationship from Oliver.

Table 7. Result

68

type(r)

"DIRECTED"

1 row

Match by relationship type

When you know the relationship type you want to match on, you can specify it by using a colontogether with the relationship type.

Query

MATCH (wallstreet:Movie { title:'Wall Street' })<-[:ACTED_IN]-(actor)RETURN actor.name

Returns all actors that ACTED_IN Wall Street.

Table 8. Result

actor.name

"Michael Douglas"

"Martin Sheen"

"Charlie Sheen"

3 rows

Match by multiple relationship types

To match on one of multiple types, you can specify this by chaining them together with the pipesymbol |.

Query

MATCH (wallstreet { title:'Wall Street' })<-[:ACTED_IN|:DIRECTED]-(person)RETURN person.name

Returns nodes with an ACTED_IN or DIRECTED relationship to Wall Street.

Table 9. Result

person.name

"Oliver Stone"

"Michael Douglas"

"Martin Sheen"

"Charlie Sheen"

4 rows

Match by relationship type and use an variable

If you both want to introduce an variable to hold the relationship, and specify the relationship typeyou want, just add them both, like this:

69

Query

MATCH (wallstreet { title:'Wall Street' })<-[r:ACTED_IN]-(actor)RETURN r.role

Returns ACTED_IN roles for Wall Street.

Table 10. Result

r.role

"Gordon Gekko"

"Carl Fox"

"Bud Fox"

3 rows

8.1.4. Relationships in depth

Inside a single pattern, relationships will only be matched once. You can read moreabout this in Uniqueness.

Relationship types with uncommon characters

Sometimes your database will have types with non-letter characters, or with spaces in them. Use `(backtick) to quote these. To demonstrate this we can add an additional relationship between CharlieSheen and Rob Reiner:

Query

MATCH (charlie:Person { name:'Charlie Sheen' }),(rob:Person { name:'Rob Reiner' })CREATE (rob)-[:`TYPEWITH SPACE`]->(charlie)

Which leads to the following graph:

Person


Movie



Person



Movie



Person




Person


DIRECTED

Person

name = 'Rob Reiner'

TYPE WITH SPACE

DIRECTED

Figure 9. Graph

Query

MATCH (n { name:'Rob Reiner' })-[r:`TYPEWITH SPACE`]->()RETURN type(r)

Returns a relationship type with a space in it

Table 11. Result

type(r)

"TYPE WITH SPACE"

70

type(r)

1 row

Multiple relationships

Relationships can be expressed by using multiple statements in the form of ()--(), or they can bestrung together, like this:

Query

MATCH (charlie { name:'Charlie Sheen' })-[:ACTED_IN]->(movie)<-[:DIRECTED]-(director)RETURN movie.title, director.name

Returns the movie Charlie acted in and its director.

Table 12. Result

movie.title director.name

"Wall Street" "Oliver Stone"

1 row

Variable length relationships

Nodes that are a variable number of relationship→node hops away can be found using the followingsyntax: -[:TYPE*minHops..maxHops]→. minHops and maxHops are optional and default to 1 andinfinity respectively. When no bounds are given the dots may be omitted. The dots may also beomitted when setting only one bound and this implies a fixed length pattern.

Query

MATCH (martin { name:'Charlie Sheen' })-[:ACTED_IN*1..3]-(movie:Movie)RETURN movie.title

Returns all movies related to Charlie by 1 to 3 hops.

Table 13. Result

movie.title

"Wall Street"



3 rows

Relationship variable in variable length relationships

When the connection between two nodes is of variable length, a relationship variable becomes a listof relationships.

Query

MATCH (actor { name:'Charlie Sheen' })-[r:ACTED_IN*2]-(co_actor)RETURN r

Returns a list of relationships.

71

Table 14. Result

r

[:ACTED_IN[0]\{role:"Bud Fox"\},:ACTED_IN[1]\{role:"Carl Fox"\}]

[:ACTED_IN[0]\{role:"Bud Fox"\},:ACTED_IN[2]\{role:"Gordon Gekko"\}]

2 rows

Match with properties on a variable length path

A variable length relationship with properties defined on in it means that all relationships in the pathmust have the property set to the given value. In this query, there are two paths between CharlieSheen and his father Martin Sheen. One of them includes a blocked'' relationship and the otherdoesn’t. In this case we first alter the original graph by using the following query to addblocked'' and ``unblocked'' relationships:

Query

MATCH (charlie:Person { name:'Charlie Sheen' }),(martin:Person { name:'Martin Sheen' })CREATE (charlie)-[:X { blocked:false }]->(:Unblocked)<-[:X { blocked:false }]-(martin)CREATE (charlie)-[:X { blocked:true }]->(:Blocked)<-[:X { blocked:false }]-(martin)

This means that we are starting out with the following graph:

Person


Blocked

Xblocked = true

Unblocked

Xblocked = false

Movie



Person


Xblocked = false

Xblocked = false


Movie



Person




Person


DIRECTED

Person

name = 'Rob Reiner'

DIRECTED

Figure 10. Graph

Query

MATCH p =(charlie:Person)-[* { blocked:false }]-(martin:Person)WHERE charlie.name = 'Charlie Sheen' AND martin.name = 'Martin Sheen'RETURN p

Returns the paths between Charlie and Martin Sheen where all relationships have the blockedproperty set to FALSE.

Table 15. Result

p

[Node[0]\{name:"CharlieSheen"\},:X[7]\{blocked:false\},Node[7]\{\},:X[8]\{blocked:false\},Node[1]\{name:"Martin Sheen"\}]

1 row

Zero length paths

Using variable length paths that have the lower bound zero means that two variables can point to thesame node. If the path length between two nodes is zero, they are by definition the same node. Notethat when matching zero length paths the result may contain a match even when matching on arelationship type not in use.

Query

MATCH (wallstreet:Movie { title:'Wall Street' })-[*0..1]-(x)RETURN x

Returns the movie itself as well as actors and directors one relationship away

72

Table 16. Result

x



Node[1]\{name:"Martin Sheen"\}

Node[2]\{name:"Michael Douglas"\}


5 rows

Named path

If you want to return or filter on a path in your pattern graph, you can a introduce a named path.

Query

MATCH p =(michael { name:'Michael Douglas' })-->()RETURN p

Returns the two paths starting from Michael

Table 17. Result

p

[Node[2]\{name:"Michael Douglas"\},:ACTED_IN[5]\{role:"President Andrew Shepherd"\},Node[6]\{title:"TheAmerican President"\}]

[Node[2]\{name:"Michael Douglas"\},:ACTED_IN[2]\{role:"Gordon Gekko"\},Node[5]\{title:"Wall Street"\}]

2 rows

Matching on a bound relationship

When your pattern contains a bound relationship, and that relationship pattern doesn’t specifydirection, Cypher will try to match the relationship in both directions.

Query

MATCH (a)-[r]-(b)WHERE id(r)= 0RETURN a,b

This returns the two connected nodes, once as the start node, and once as the end node

Table 18. Result

a b

Node[0]\{name:"Charlie Sheen"\} Node[5]\{title:"Wall Street"\}

Node[5]\{title:"Wall Street"\} Node[0]\{name:"Charlie Sheen"\}

2 rows

8.1.5. Shortest path

73

Single shortest path

Finding a single shortest path between two nodes is as easy as using the shortestPath function. It’sdone like this:

Query

MATCH (martin:Person { name:"Martin Sheen" }),(oliver:Person { name:"Oliver Stone" }), p =shortestPath((martin)-[*..15]-(oliver))RETURN p

This means: find a single shortest path between two nodes, as long as the path is max 15relationships long. Inside of the parentheses you define a single link of a path — the starting node, theconnecting relationship and the end node. Characteristics describing the relationship like relationshiptype, max hops and direction are all used when finding the shortest path. If there is a WHERE clausefollowing the match of a shortestPath, relevant predicates will be included in the shortestPath. If thepredicate is a NONE() or ALL() on the relationship elements of the path, it will be used during thesearch to improve performance (see Shortest path planning).

Table 19. Result

p

[Node[1]\{name:"Martin Sheen"\},:ACTED_IN[1]\{role:"Carl Fox"\},Node[5]\{title:"WallStreet"\},:DIRECTED[3]\{\},Node[3]\{name:"Oliver Stone"\}]

1 row

Single shortest path with predicates

Predicates used in the WHERE clause that apply to the shortest path pattern are evaluated beforedeciding what the shortest matching path is.

Query

MATCH (charlie:Person { name:"Charlie Sheen" }),(martin:Person { name:"Martin Sheen" }), p =shortestPath((charlie)-[*]-(martin))WHERE NONE (r IN rels(p) WHERE type(r)= "FATHER")RETURN p

This query will find the shortest path between 'Charlie Sheen' and 'Martin Sheen', and the WHEREpredicate will ensure that we don’t consider the father/son relationship between the two.

Table 20. Result

p

[Node[0]\{name:"Charlie Sheen"\},:ACTED_IN[0]\{role:"Bud Fox"\},Node[5]\{title:"WallStreet"\},:ACTED_IN[1]\{role:"Carl Fox"\},Node[1]\{name:"Martin Sheen"\}]

1 row

All shortest paths

Finds all the shortest paths between two nodes.

Query

MATCH (martin:Person { name:"Martin Sheen" }),(michael:Person { name:"Michael Douglas" }), p =allShortestPaths((martin)-[*]-(michael))RETURN p

74

Finds the two shortest paths between 'Martin' and 'Michael'.

Table 21. Result

p

[Node[1]\{name:"Martin Sheen"\},:ACTED_IN[1]\{role:"Carl Fox"\},Node[5]\{title:"WallStreet"\},:ACTED_IN[2]\{role:"Gordon Gekko"\},Node[2]\{name:"Michael Douglas"\}]

[Node[1]\{name:"Martin Sheen"\},:ACTED_IN[4]\{role:"A.J. MacInerney"\},Node[6]\{title:"The AmericanPresident"\},:ACTED_IN[5]\{role:"President Andrew Shepherd"\},Node[2]\{name:"Michael Douglas"\}]

2 rows

8.1.6. Get node or relationship by id

Node by id

Searching for nodes by id can be done with the id() function in a predicate.

Neo4j reuses its internal ids when nodes and relationships are deleted. This meansthat applications using, and relying on internal Neo4j ids, are brittle or at risk ofmaking mistakes. It is therefor recommended to rather use application generatedids.

Query

MATCH (n)WHERE id(n)= 0RETURN n

The corresponding node is returned.

Table 22. Result

n


1 row

Relationship by id

Search for relationships by id can be done with the id() function in a predicate.

This is not recommended practice. See Node by id for more information on the use of Neo4j ids.

Query

MATCH ()-[r]->()WHERE id(r)= 0RETURN r

The relationship with id 0 is returned.

Table 23. Result

r

:ACTED_IN[0]\{role:"Bud Fox"\}

1 row

75

Multiple nodes by id

Multiple nodes are selected by specifying them in an IN clause.

Query

MATCH (n)WHERE id(n) IN [0,3,5]RETURN n

This returns the nodes listed in the IN expression.

Table 24. Result

n




3 rows

8.2. Optional Match

The OPTIONAL MATCH clause is used to search for the pattern described in it, while using

NULLs for missing parts of the pattern.

8.2.1. Introduction

OPTIONAL MATCH matches patterns against your graph database, just like MATCH does. Thedifference is that if no matches are found, OPTIONAL MATCH will use NULLs for missing parts of thepattern. OPTIONAL MATCH could be considered the Cypher equivalent of the outer join in SQL.

Either the whole pattern is matched, or nothing is matched. Remember that WHERE is part of thepattern description, and the predicates will be considered while looking for matches, not after. Thismatters especially in the case of multiple (OPTIONAL) MATCH clauses, where it is crucial to put WHEREtogether with the MATCH it belongs to.

To understand the patterns used in the OPTIONAL MATCH clause, read Patterns.


Movie

name = 'WallStreet'title = 'Wall Street'

Person

name = 'Rob Reiner'

Movie

title = 'The American President'name = 'TheAmericanPresident'

DIRECTED

Person


DIRECTED

Person


ACTED_INPerson


FATHER

ACTED_IN ACTED_IN

Person


ACTED_IN ACTED_IN

Figure 11. Graph

76

8.2.2. Relationship

If a relationship is optional, use the OPTIONAL MATCH clause. This is similar to how a SQL outer joinworks. If the relationship is there, it is returned. If it’s not, NULL is returned in it’s place.

Query

MATCH (a:Movie { title: 'Wall Street' })OPTIONAL MATCH (a)-->(x)RETURN x

Returns NULL, since the node has no outgoing relationships.

Result

+--------+| x |+--------+| <null> |+--------+1 row

8.2.3. Properties on optional elements

Returning a property from an optional element that is NULL will also return NULL.

Query

MATCH (a:Movie { title: 'Wall Street' })OPTIONAL MATCH (a)-->(x)RETURN x, x.name

Returns the element x (NULL in this query), and NULL as its name.

Result

+-----------------+| x | x.name |+-----------------+| <null> | <null> |+-----------------+1 row

8.2.4. Optional typed and named relationship

Just as with a normal relationship, you can decide which variable it goes into, and what relationshiptype you need.

Query

MATCH (a:Movie { title: 'Wall Street' })OPTIONAL MATCH (a)-[r:ACTS_IN]->()RETURN r

This returns a node, and NULL, since the node has no outgoing ACTS_IN relationships.

77

Result

+--------+| r |+--------+| <null> |+--------+1 row

8.3. Where

WHERE adds constraints to the patterns in a MATCH or OPTIONAL MATCH clause or filters the

results of a WITH clause.

WHERE is not a clause in it’s own right — rather, it’s part of MATCH, OPTIONAL MATCH, START and WITH.

In the case of WITH and START, WHERE simply filters the results.

For MATCH and OPTIONAL MATCH on the other hand, WHERE adds constraints to the patterns described. Itshould not be seen as a filter after the matching is finished.

In the case of multiple MATCH / OPTIONAL MATCH clauses, the predicate in WHERE isalways a part of the patterns in the directly preceding MATCH / OPTIONAL MATCH. Bothresults and performance may be impacted if the WHERE is put inside the wrong MATCHclause.

Swedish

name = 'Andres'age = 36belt = 'white'

email = '[email protected]'name = 'Peter'age = 34

KNOWSsince = 1999

address = 'Sweden/Malmo'name = 'Tobias'age = 25

KNOWSsince = 2012

Figure 12. Graph

8.3.1. Basic usage

Boolean operations

You can use the expected boolean operators AND and OR, and also the boolean function NOT. SeeWorking with NULL for more information on how this works with NULL.

Query

MATCH (n)WHERE n.name = 'Peter' XOR (n.age < 30 AND n.name = "Tobias") OR NOT (n.name = "Tobias" OR n.name="Peter")RETURN n

78

Result

+----------------------------------------------------------+| n |+----------------------------------------------------------+| Node[0]{name:"Andres",age:36,belt:"white"} || Node[1]{address:"Sweden/Malmo",name:"Tobias",age:25} || Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+3 rows

Filter on node label

To filter nodes by label, write a label predicate after the WHERE keyword using WHERE n:foo.

Query

MATCH (n)WHERE n:SwedishRETURN n

The "Andres" node will be returned.

Result

+--------------------------------------------+| n |+--------------------------------------------+| Node[0]{name:"Andres",age:36,belt:"white"} |+--------------------------------------------+1 row

Filter on node property

To filter on a node property, write your clause after the WHERE keyword.

Query

MATCH (n)WHERE n.age < 30RETURN n

"Tobias" is returned because he is younger than 30.

Result

+------------------------------------------------------+| n |+------------------------------------------------------+| Node[1]{address:"Sweden/Malmo",name:"Tobias",age:25} |+------------------------------------------------------+1 row

Filter on relationship property

To filter on a relationship property, write your clause after the WHERE keyword.

79

Query

MATCH (n)-[k:KNOWS]->(f)WHERE k.since < 2000RETURN f

"Peter" is returned because Andres knows him since before 2000.

Result

+----------------------------------------------------------+| f |+----------------------------------------------------------+| Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+1 row

Filter on dynamic node property

To filter on a property using a dynamically computed name, use square bracket syntax.

Parameters

{ "prop" : "AGE"}

Query

MATCH (n)WHERE n[toLower({ prop })]< 30RETURN n

"Tobias" is returned because he is younger than 30.

Result


Property exists

Use the EXISTS() function to only include nodes or relationships in which a property exists.

Query

MATCH (n)WHERE exists(n.belt)RETURN n

"Andres" will be returned because he is the only one with a belt property.

The HAS() function has been superseded by EXISTS() and has been removed.

80

Result


8.3.2. String matching

The start and end of strings can be matched using STARTS WITH and ENDS WITH. To match regardless oflocation in a string, use CONTAINS. The matching is case-sensitive.

Match the start of a string

The STARTS WITH operator is used to perform case-sensitive matching on the start of strings.

Query

MATCH (n)WHERE n.name STARTS WITH 'Pet'RETURN n

"Peter" will be returned because his name starts with Pet.

Result

+----------------------------------------------------------+| n |+----------------------------------------------------------+| Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+1 row

Match the end of a string

The ENDS WITH operator is used to perform case-sensitive matching on the end of strings.

Query

MATCH (n)WHERE n.name ENDS WITH 'ter'RETURN n

"Peter" will be returned because his name ends with ter.

Result


Match anywhere in a string

The CONTAINS operator is used to perform case-sensitive matching regardless of location in strings.

81

Query

MATCH (n)WHERE n.name CONTAINS 'ete'RETURN n

"Peter" will be returned because his name contains ete.

Result


String matching negation

Use the NOT keyword to exclude all matches on given string from your result:

Query

MATCH (n)WHERE NOT n.name ENDS WITH 's'RETURN n

"Peter" will be returned because his name does not end with s.

Result


8.3.3. Regular expressions

Cypher supports filtering using regular expressions. The regular expression syntax is inherited fromthe Java regular expressions (https://docs.oracle.com/javase/7/docs/api/java/util/regex/Pattern.html). Thisincludes support for flags that change how strings are matched, including case-insensitive (?i),multiline (?m) and dotall (?s). Flags are given at the start of the regular expression, for example MATCH(n) WHERE n.name =~ '(?i)Lon.*' RETURN n will return nodes with name London or with name LonDoN.

Regular expressions

You can match on regular expressions by using =~ "regexp", like this:

Query

MATCH (n)WHERE n.name =~ 'Tob.*'RETURN n

"Tobias" is returned because his name starts with Tob.

82

https://docs.oracle.com/javase/7/docs/api/java/util/regex/Pattern.html

Result


Escaping in regular expressions

If you need a forward slash inside of your regular expression, escape it. Remember that back slashneeds to be escaped in string literals.

Query

MATCH (n)WHERE n.address =~ 'Sweden\\/Malmo'RETURN n

"Tobias" is returned because his address is in Sweden/Malmo.

Result


Case insensitive regular expressions

By pre-pending a regular expression with (?i), the whole expression becomes case insensitive.

Query

MATCH (n)WHERE n.name =~ '(?i)ANDR.*'RETURN n

"Andres" is returned because his name starts with ANDR regardless of case.

Result


8.3.4. Using path patterns in WHERE

Filter on patterns

Patterns are expressions in Cypher, expressions that return a list of paths. List expressions are alsopredicates — an empty list represents false, and a non-empty represents true.

So, patterns are not only expressions, they are also predicates. The only limitation to your pattern is

83

that you must be able to express it in a single path. You can not use commas between multiple pathslike you do in MATCH. You can achieve the same effect by combining multiple patterns with AND.

Note that you can not introduce new variables here. Although it might look very similar to the MATCHpatterns, the WHERE clause is all about eliminating matched subgraphs. MATCH (a)-[]→(b) is verydifferent from WHERE (a)-[]→(b); the first will produce a subgraph for every path it can find betweena and b, and the latter will eliminate any matched subgraphs where a and b do not have a directedrelationship chain between them.

Query

MATCH (tobias { name: 'Tobias' }),(others)WHERE others.name IN ['Andres', 'Peter'] AND (tobias)<--(others)RETURN others

Nodes that have an outgoing relationship to the "Tobias" node are returned.

Result

+--------------------------------------------+| others |+--------------------------------------------+| Node[0]{name:"Andres",age:36,belt:"white"} |+--------------------------------------------+1 row

Filter on patterns using NOT

The NOT function can be used to exclude a pattern.

Query

MATCH (persons),(peter { name: 'Peter' })WHERE NOT (persons)-->(peter)RETURN persons

Nodes that do not have an outgoing relationship to the "Peter" node are returned.

Result

+----------------------------------------------------------+| persons |+----------------------------------------------------------+| Node[1]{address:"Sweden/Malmo",name:"Tobias",age:25} || Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+2 rows

Filter on patterns with properties

You can also add properties to your patterns:

Query

MATCH (n)WHERE (n)-[:KNOWS]-({ name:'Tobias' })RETURN n

Finds all nodes that have a KNOWS relationship to a node with the name "Tobias".

84

Result


Filtering on relationship type

You can put the exact relationship type in the MATCH pattern, but sometimes you want to be able to domore advanced filtering on the type. You can use the special property TYPE to compare the type withsomething else. In this example, the query does a regular expression comparison with the name ofthe relationship type.

Query

MATCH (n)-[r]->()WHERE n.name='Andres' AND type(r)=~ 'K.*'RETURN r

This returns relationships that has a type whose name starts with K.

Result

+-----------------------+| r |+-----------------------+| :KNOWS[1]{since:1999} || :KNOWS[0]{since:2012} |+-----------------------+2 rows

8.3.5. Lists

IN operator

To check if an element exists in a list, you can use the IN operator.

Query

MATCH (a)WHERE a.name IN ["Peter", "Tobias"]RETURN a

This query shows how to check if a property exists in a literal list.

Result

+----------------------------------------------------------+| a |+----------------------------------------------------------+| Node[1]{address:"Sweden/Malmo",name:"Tobias",age:25} || Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+2 rows

85

8.3.6. Missing properties and values

Default to false if property is missing

As missing properties evaluate to NULL, the comparision in the example will evaluate to FALSE fornodes without the belt property.

Query

MATCH (n)WHERE n.belt = 'white'RETURN n

Only nodes with white belts are returned.

Result


Default to true if property is missing

If you want to compare a property on a graph element, but only if it exists, you can compare theproperty against both the value you are looking for and NULL, like:

Query

MATCH (n)WHERE n.belt = 'white' OR n.belt IS NULL RETURN nORDER BY n.name

This returns all nodes, even those without the belt property.

Result

+----------------------------------------------------------+| n |+----------------------------------------------------------+| Node[0]{name:"Andres",age:36,belt:"white"} || Node[2]{email:"[email protected]",name:"Peter",age:34} || Node[1]{address:"Sweden/Malmo",name:"Tobias",age:25} |+----------------------------------------------------------+3 rows

Filter on NULL

Sometimes you might want to test if a value or a variable is NULL. This is done just like SQL does it,with IS NULL. Also like SQL, the negative is IS NOT NULL, although NOT(IS NULL x) also works.

Query

MATCH (person)WHERE person.name = 'Peter' AND person.belt IS NULL RETURN person

Nodes that have name Peter but no belt property are returned.

86

Result

+----------------------------------------------------------+| person |+----------------------------------------------------------+| Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+1 row

8.3.7. Using ranges

Simple range

To check for an element being inside a specific range, use the inequality operators <, ⇐, >=, >.

Query

MATCH (a)WHERE a.name >= 'Peter'RETURN a

Nodes having a name property lexicographically greater than or equal to 'Peter' are returned.

Result

+----------------------------------------------------------+| a |+----------------------------------------------------------+| Node[1]{address:"Sweden/Malmo",name:"Tobias",age:25} || Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+2 rows

Composite range

Several inequalities can be used to construct a range.

Query

MATCH (a)WHERE a.name > 'Andres' AND a.name < 'Tobias'RETURN a

Nodes having a name property lexicographically between 'Andres' and 'Tobias' are returned.

Result

+----------------------------------------------------------+| a |+----------------------------------------------------------+| Node[2]{email:"[email protected]",name:"Peter",age:34} |+----------------------------------------------------------+1 row

8.4. Start

Find starting points through legacy indexes.

87

The START clause should only be used when accessing legacy indexes (see[indexing]). In all other cases, use MATCH instead (see Match).

In Cypher, every query describes a pattern, and in that pattern one can have multiple starting points.A starting point is a relationship or a node where a pattern is anchored. Using START you can onlyintroduce starting points by legacy index seeks. Note that trying to use a legacy index that doesn’texist will generate an error.

This is the graph the examples are using:

Node[0]

name = 'A'

Node[2]

name = 'C'

KNOWS

Node[1]

name = 'B'

KNOWS

Figure 13. Graph

8.4.1. Get node or relationship from index

Node by index seek

When the starting point can be found by using index seeks, it can be done like this: node:index-name(key = "value"). In this example, there exists a node index named nodes.

Query

START n=node:nodes(name = "A")RETURN n

The query returns the node indexed with the name "A".

Result

+-------------------+| n |+-------------------+| Node[0]{name:"A"} |+-------------------+1 row

Relationship by index seek

When the starting point can be found by using index seeks, it can be done like this:relationship:index-name(key = "value").

Query

START r=relationship:rels(name = "Andrés")RETURN r

The relationship indexed with the name property set to "Andrés" is returned by the query.

88

Result

+--------------------------+| r |+--------------------------+| :KNOWS[0]{name:"Andrés"} |+--------------------------+1 row

Node by index query

When the starting point can be found by more complex Lucene queries, this is the syntax to use:node:index-name("query").This allows you to write more advanced index queries.

Query

START n=node:nodes("name:A")RETURN n

The node indexed with name "A" is returned by the query.

Result

+-------------------+| n |+-------------------+| Node[0]{name:"A"} |+-------------------+1 row

8.5. Aggregation

8.5.1. Introduction

To calculate aggregated data, Cypher offers aggregation, much like SQL’s GROUP BY.

Aggregate functions take multiple input values and calculate an aggregated value from them.Examples are avg that calculates the average of multiple numeric values, or min that finds thesmallest numeric value in a set of values.

Aggregation can be done over all the matching subgraphs, or it can be further divided by introducingkey values. These are non-aggregate expressions, that are used to group the values going into theaggregate functions.

So, if the return statement looks something like this:

RETURN n, count(*)

We have two return expressions: n, and count(). The first, n, is no aggregate function, and so itwill be the grouping key. The latter, count() is an aggregate expression. So the matchingsubgraphs will be divided into different buckets, depending on the grouping key. The aggregatefunction will then run on these buckets, calculating the aggregate values.

If you want to use aggregations to sort your result set, the aggregation must be included in theRETURN to be used in your ORDER BY.

The last piece of the puzzle is the DISTINCT keyword. It is used to make all values unique beforerunning them through an aggregate function.

89

An example might be helpful. In this case, we are running the query against the following data:

Person

name = 'A'property = 13

Person

name = 'D'eyes = 'brown'

KNOWS

Person

name = 'C'property = 44eyes = 'blue'

KNOWS

Person

name = 'B'property = 33eyes = 'blue'

KNOWS

Person

name = 'D'

KNOWS KNOWS

Query

MATCH (me:Person)-->(friend:Person)-->(friend_of_friend:Person)WHERE me.name = 'A'RETURN count(DISTINCT friend_of_friend), count(friend_of_friend)

In this example we are trying to find all our friends of friends, and count them. The first aggregatefunction, count(DISTINCT friend_of_friend), will only see a friend_of_friend once — DISTINCT removesthe duplicates. The latter aggregate function, count(friend_of_friend), might very well see the samefriend_of_friend multiple times. In this case, both B and C know D and thus D will get counted twice,when not using DISTINCT.

Result

+------------------------------------------------------------+| count(distinct friend_of_friend) | count(friend_of_friend) |+------------------------------------------------------------+| 1 | 2 |+------------------------------------------------------------+1 row

The following examples are assuming the example graph structure below.

Person

name = 'A'property = 13

Person

name = 'D'eyes = 'brown'

KNOWS

Person

name = 'C'property = 44eyes = 'blue'

KNOWS

Person

name = 'B'property = 33eyes = 'blue'

KNOWS

Figure 14. Graph

8.5.2. COUNT

COUNT is used to count the number of rows.

COUNT can be used in two forms — COUNT(*) which just counts the number of matching rows, andCOUNT(<expression>), which counts the number of non-NULL values in <expression>.

90

Count nodes

To count the number of nodes, for example the number of nodes connected to one node, you can usecount(*).

Query

MATCH (n { name: 'A' })-->(x)RETURN n, count(*)

This returns the start node and the count of related nodes.

Result

+------------------------------------------+| n | count(*) |+------------------------------------------+| Node[0]{name:"A",property:13} | 3 |+------------------------------------------+1 row

Group Count Relationship Types

To count the groups of relationship types, return the types and count them with count(*).

Query

MATCH (n { name: 'A' })-[r]->()RETURN type(r), count(*)

The relationship types and their group count is returned by the query.

Result

+--------------------+| type(r) | count(*) |+--------------------+| "KNOWS" | 3 |+--------------------+1 row

Count entities

Instead of counting the number of results with count(*), it might be more expressive to include thename of the variable you care about.

Query

MATCH (n { name: 'A' })-->(x)RETURN count(x)

The example query returns the number of connected nodes from the start node.

Result

+----------+| count(x) |+----------+| 3 |+----------+1 row

91

Count non-null values

You can count the non-NULL values by using count(<expression>).

Query

MATCH (n:Person)RETURN count(n.property)

The count of related nodes with the property property set is returned by the query.

Result

+-------------------+| count(n.property) |+-------------------+| 3 |+-------------------+1 row

8.5.3. Statistics

sum

The sum aggregation function simply sums all the numeric values it encounters. NULLs are silentlydropped.

Query

MATCH (n:Person)RETURN sum(n.property)

This returns the sum of all the values in the property property.

Result

+-----------------+| sum(n.property) |+-----------------+| 90 |+-----------------+1 row

avg

avg calculates the average of a numeric column.

Query

MATCH (n:Person)RETURN avg(n.property)

The average of all the values in the property property is returned by the example query.

92

Result

+-----------------+| avg(n.property) |+-----------------+| 30.0 |+-----------------+1 row

percentileDisc

percentileDisc calculates the percentile of a given value over a group, with a percentile from 0.0 to 1.0.It uses a rounding method, returning the nearest value to the percentile. For interpolated values, seepercentileCont.

Query

MATCH (n:Person)RETURN percentileDisc(n.property, 0.5)

The 50th percentile of the values in the property property is returned by the example query. In thiscase, 0.5 is the median, or 50th percentile.

Result

+---------------------------------+| percentileDisc(n.property, 0.5) |+---------------------------------+| 33 |+---------------------------------+1 row

percentileCont

percentileCont calculates the percentile of a given value over a group, with a percentile from 0.0 to1.0. It uses a linear interpolation method, calculating a weighted average between two values, if thedesired percentile lies between them. For nearest values using a rounding method, see percentileDisc.

Query

MATCH (n:Person)RETURN percentileCont(n.property, 0.4)

The 40th percentile of the values in the property property is returned by the example query,calculated with a weighted average.

Result

+---------------------------------+| percentileCont(n.property, 0.4) |+---------------------------------+| 29.0 |+---------------------------------+1 row

stdev

stdev calculates the standard deviation for a given value over a group. It uses a standard two-passmethod, with N - 1 as the denominator, and should be used when taking a sample of the populationfor an unbiased estimate. When the standard variation of the entire population is being calculated,

93

stdevp should be used.

Query

MATCH (n)WHERE n.name IN ['A', 'B', 'C']RETURN stdev(n.property)

The standard deviation of the values in the property property is returned by the example query.

Result

+--------------------+| stdev(n.property) |+--------------------+| 15.716233645501712 |+--------------------+1 row

stdevp

stdevp calculates the standard deviation for a given value over a group. It uses a standard two-passmethod, with N as the denominator, and should be used when calculating the standard deviation foran entire population. When the standard variation of only a sample of the population is beingcalculated, stdev should be used.

Query

MATCH (n)WHERE n.name IN ['A', 'B', 'C']RETURN stdevp(n.property)

The population standard deviation of the values in the property property is returned by the examplequery.

Result

+--------------------+| stdevp(n.property) |+--------------------+| 12.832251036613439 |+--------------------+1 row

max

max find the largest value in a numeric column.

Query

MATCH (n:Person)RETURN max(n.property)

The largest of all the values in the property property is returned.

94

Result

+-----------------+| max(n.property) |+-----------------+| 44 |+-----------------+1 row

min

min takes a numeric property as input, and returns the smallest value in that column.

Query

MATCH (n:Person)RETURN min(n.property)

This returns the smallest of all the values in the property property.

Result

+-----------------+| min(n.property) |+-----------------+| 13 |+-----------------+1 row

8.5.4. collect

collect collects all the values into a list. It will ignore NULLs.

Query

MATCH (n:Person)RETURN collect(n.property)

Returns a single row, with all the values collected.

Result

+---------------------+| collect(n.property) |+---------------------+| [13,33,44] |+---------------------+1 row

8.5.5. DISTINCT

All aggregation functions also take the DISTINCT modifier, which removes duplicates from the values.So, to count the number of unique eye colors from nodes related to a, this query can be used:

Query

MATCH (a:Person { name: 'A' })-->(b)RETURN count(DISTINCT b.eyes)

Returns the number of eye colors.

95

Result

+------------------------+| count(distinct b.eyes) |+------------------------+| 2 |+------------------------+1 row

8.6. Load CSV

LOAD CSV is used to import data from CSV files.

• The URL of the CSV file is specified by using FROM followed by an arbitrary expression evaluating tothe URL in question.

• It is required to specify a variable for the CSV data using AS.

• LOAD CSV supports resources compressed with gzip, Deflate, as well as ZIP archives.

• CSV files can be stored on the database server and are then accessible using a file:/// URL.Alternatively, LOAD CSV also supports accessing CSV files via HTTPS, HTTP, and FTP.

• LOAD CSV will follow HTTP redirects but for security reasons it will not follow redirects that changesthe protocol, for example if the redirect is going from HTTPS to HTTP.

96

Configuration settings for file URLs

[config_dbms.security.allow_csv_import_from_file_urls]

This setting determines if Cypher will allow the use of file:/// URLs whenloading data using LOAD CSV. Such URLs identify files on the filesystem of thedatabase server. Default is true.

[config_dbms.directories.import]

Sets the root directory for file:/// URLs used with the Cypher LOAD CSV clause.This must be set to a single directory on the filesystem of the database server,and will make all requests to load from file:/// URLs relative to the specifieddirectory (similar to how a unix chroot operates). By default, this setting is notconfigured.

• When not set, file URLs will be resolved as relative to the root of thedatabase server filesystem. If this is the case, a file URL will typically look likefile:///home/username/myfile.csv or file:///C:/Users/username/myfile.csv.Using these URLs in LOAD CSV will read content from files on the databaseserver filesystem, specifically /home/username/myfile.csv andC:\Users\username\myfile.csv respectively. For security reasons you may notwant users to be able to load files located anywhere on the database serverfilesystem and should set dbms.directories.import to a safe directory to loadfiles from.

• When set, file URLs will be resolved as relative to the directory it’s set to. Inthis case a file URL will typically look like file:///myfile.csv orfile:///myproject/myfile.csv.

• If set to import using the above URLs in LOAD CSV would read contentfrom import/myfile.csv and import/myproject/myfile.csv respectively,where both are relative to the neo4j root directory.

• If set to /home/neo4j using the above URLs in LOAD CSV would readcontent from /home/neo4j/myfile.csv and/home/neo4j/myproject/myfile.csv respectively.

See the examples below for further details.

There is also a worked example, see Importing CSV files with Cypher.

8.6.1. CSV file format

The CSV file to use with LOAD CSV must have the following characteristics:

• the character encoding is UTF-8;

• the end line termination is system dependent, e.g., it is \n on unix or \r\n on windows;

• the default field terminator is ,;

• the field terminator character can be change by using the option FIELDTERMINATOR available inthe LOAD CSV command;

• quoted strings are allowed in the CSV file and the quotes are dropped when reading the data;

• the character for string quotation is double quote ";

• the escape character is \.

97

8.6.2. Import data from a CSV file

To import data from a CSV file into Neo4j, you can use LOAD CSV to get the data into your query. Thenyou write it to your database using the normal updating clauses of Cypher.

artists.csv

"1","ABBA","1992""2","Roxette","1986""3","Europe","1979""4","The Cardigans","1992"

Query

LOAD CSV FROM 'http://neo4j.com/docs/3.0.1/csv/artists.csv' AS lineCREATE (:Artist { name: line[1], year: toInt(line[2])})

A new node with the Artist label is created for each row in the CSV file. In addition, two columns fromthe CSV file are set as properties on the nodes.

Result


8.6.3. Import data from a CSV file containing headers

When your CSV file has headers, you can view each row in the file as a map instead of as an array ofstrings.

artists-with-headers.csv

"Id","Name","Year""1","ABBA","1992""2","Roxette","1986""3","Europe","1979""4","The Cardigans","1992"

Query

LOAD CSV WITH HEADERS FROM 'http://neo4j.com/docs/3.0.1/csv/artists-with-headers.csv' AS lineCREATE (:Artist { name: line.Name, year: toInt(line.Year)})

This time, the file starts with a single row containing column names. Indicate this using WITH HEADERSand you can access specific fields by their corresponding column name.

Result


98

8.6.4. Import data from a CSV file with a custom field delimiter

Sometimes, your CSV file has other field delimiters than commas. You can specify which delimiteryour file uses using FIELDTERMINATOR.

artists-fieldterminator.csv

"1";"ABBA";"1992""2";"Roxette";"1986""3";"Europe";"1979""4";"The Cardigans";"1992"

Query

LOAD CSV FROM 'http://neo4j.com/docs/3.0.1/csv/artists-fieldterminator.csv' AS line FIELDTERMINATOR ';'CREATE (:Artist { name: line[1], year: toInt(line[2])})

As values in this file are separated by a semicolon, a custom FIELDTERMINATOR is specified in theLOAD CSV clause.

Result


8.6.5. Importing large amounts of data

If the CSV file contains a significant number of rows (approaching hundreds of thousands or millions),USING PERIODIC COMMIT can be used to instruct Neo4j to perform a commit after a number of rows.This reduces the memory overhead of the transaction state. By default, the commit will happen every1000 rows. For more information, see Using Periodic Commit.

Query

USING PERIODIC COMMITLOAD CSV FROM 'http://neo4j.com/docs/3.0.1/csv/artists.csv' AS lineCREATE (:Artist { name: line[1], year: toInt(line[2])})

Result


8.6.6. Setting the rate of periodic commits

You can set the number of rows as in the example, where it is set to 500 rows.

99

Query

USING PERIODIC COMMIT 500LOAD CSV FROM 'http://neo4j.com/docs/3.0.1/csv/artists.csv' AS lineCREATE (:Artist { name: line[1], year: toInt(line[2])})

Result


8.6.7. Import data containing escaped characters

In this example, we both have additional quotes around the values, as well as escaped quotes insideone value.

artists-with-escaped-char.csv

"1","The ""Symbol""","1992"

Query

LOAD CSV FROM 'http://neo4j.com/docs/3.0.1/csv/artists-with-escaped-char.csv' AS lineCREATE (a:Artist { name: line[1], year: toInt(line[2])})RETURN a.name AS name, a.year AS year, length(a.name) AS length

Note that strings are wrapped in quotes in the output here. You can see that when comparing to thelength of the string in this case!

Result

+--------------------------------+| name | year | length |+--------------------------------+| "The "Symbol"" | 1992 | 12 |+--------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

100

Chapter 9. Writing ClausesWrite data to the database.

9.1. Create

The CREATE clause is used to create graph elements — nodes and relationships.

In the CREATE clause, patterns are used a lot. Read Patterns for an introduction.

9.1.1. Create nodes

Create single node

Creating a single node is done by issuing the following query.

Query

CREATE (n)

Nothing is returned from this query, except the count of affected nodes.

Result

+-------------------+| No data returned. |+-------------------+Nodes created: 1

Create multiple nodes

Creating multiple nodes is done by separating them with a comma.

Query

CREATE (n),(m)

Result

+-------------------+| No data returned. |+-------------------+Nodes created: 2

Create a node with a label

To add a label when creating a node, use the syntax below.

Query

CREATE (n:Person)

Nothing is returned from this query.

101

Result

+-------------------+| No data returned. |+-------------------+Nodes created: 1Labels added: 1

Create a node with multiple labels

To add labels when creating a node, use the syntax below. In this case, we add two labels.

Query

CREATE (n:Person:Swedish)


Result

+-------------------+| No data returned. |+-------------------+Nodes created: 1Labels added: 2

Create node and add labels and properties

When creating a new node with labels, you can add properties at the same time.

Query

CREATE (n:Person { name : 'Andres', title : 'Developer' })


Result


Return created node

Creating a single node is done by issuing the following query.

Query

CREATE (a { name : 'Andres' })RETURN a

The newly created node is returned.

102

Result

+------------------------+| a |+------------------------+| Node[0]{name:"Andres"} |+------------------------+1 rowNodes created: 1Properties set: 1

9.1.2. Create relationships

Create a relationship between two nodes

To create a relationship between two nodes, we first get the two nodes. Once the nodes are loaded,we simply create a relationship between them.

Query

MATCH (a:Person),(b:Person)WHERE a.name = 'Node A' AND b.name = 'Node B'CREATE (a)-[r:RELTYPE]->(b)RETURN r

The created relationship is returned by the query.

Result

+---------------+| r |+---------------+| :RELTYPE[0]{} |+---------------+1 rowRelationships created: 1

Create a relationship and set properties

Setting properties on relationships is done in a similar manner to how it’s done when creating nodes.Note that the values can be any expression.

Query

MATCH (a:Person),(b:Person)WHERE a.name = 'Node A' AND b.name = 'Node B'CREATE (a)-[r:RELTYPE { name : a.name + '<->' + b.name }]->(b)RETURN r

The newly created relationship is returned by the example query.

Result

+-------------------------------------+| r |+-------------------------------------+| :RELTYPE[0]{name:"Node A<->Node B"} |+-------------------------------------+1 rowRelationships created: 1Properties set: 1

103

9.1.3. Create a full path

When you use CREATE and a pattern, all parts of the pattern that are not already in scope at this timewill be created.

Query

CREATE p =(andres { name:'Andres' })-[:WORKS_AT]->(neo)<-[:WORKS_AT]-(michael { name:'Michael' })RETURN p

This query creates three nodes and two relationships in one go, assigns it to a path variable, andreturns it.

Result

+------------------------------------------------------------------------------------------+| p |+------------------------------------------------------------------------------------------+| [Node[0]{name:"Andres"},:WORKS_AT[0]{},Node[1]{},:WORKS_AT[1]{},Node[2]{name:"Michael"}] |+------------------------------------------------------------------------------------------+1 rowNodes created: 3Relationships created: 2Properties set: 2

9.1.4. Use parameters with CREATE

Create node with a parameter for the properties

You can also create a graph entity from a map. All the key/value pairs in the map will be set asproperties on the created relationship or node. In this case we add a Person label to the node as well.

Parameters


Query

CREATE (n:Person { props })RETURN n

Result

+---------------------------------------------+| n |+---------------------------------------------+| Node[0]{name:"Andres",position:"Developer"} |+---------------------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

Create multiple nodes with a parameter for their properties

By providing Cypher an array of maps, it will create a node for each map.

104

Parameters

{ "props" : [ { "name" : "Andres", "position" : "Developer" }, { "name" : "Michael", "position" : "Developer" } ]}

Query

UNWIND { props } AS mapCREATE (n)SET n = map

Result

+-------------------+| No data returned. |+-------------------+Nodes created: 2Properties set: 4

9.2. Merge

The MERGE clause ensures that a pattern exists in the graph. Either the pattern already exists,

or it needs to be created.

9.2.1. Introduction

MERGE either matches existing nodes and binds them, or it creates new data and binds that. It’s like acombination of MATCH and CREATE that additionally allows you to specify what happens if the data wasmatched or created.

For example, you can specify that the graph must contain a node for a user with a certain name. Ifthere isn’t a node with the correct name, a new node will be created and its name property set.

When using MERGE on full patterns, the behavior is that either the whole pattern matches, or the wholepattern is created. MERGE will not partially use existing patterns — it’s all or nothing. If partial matchesare needed, this can be accomplished by splitting a pattern up into multiple MERGE clauses.

As with MATCH, MERGE can match multiple occurrences of a pattern. If there are multiple matches, theywill all be passed on to later stages of the query.

The last part of MERGE is the ON CREATE and ON MATCH. These allow a query to express additionalchanges to the properties of a node or relationship, depending on if the element was MATCHed in thedatabase or if it was CREATEd.

The rule planner (see How are queries executed?) expands a MERGE pattern from the end point that hasthe variable with the lowest lexicographical order. This means that it might choose a suboptimalexpansion path, expanding from a node with a higher degree. The pattern MERGE (a:A)-[:R]→(b:B) willalways expand from a to b, so if it is known that b nodes are a better choice for start point, renamingvariables could improve performance.


105

Movie

name = 'WallStreet'title = 'Wall Street'

Person

chauffeurName = 'Ted Green'name = 'Rob Reiner'bornIn = 'New York'

Movie

title = 'The American President'name = 'TheAmericanPresident'

DIRECTED

Person

chauffeurName = 'Bill White'name = 'Oliver Stone'bornIn = 'New York'

DIRECTED

Person

chauffeurName = 'John Brown'name = 'Charlie Sheen'bornIn = 'New York'

ACTED_IN

Person

chauffeurName = 'Bob Brown'name = 'Martin Sheen'bornIn = 'Ohio'

FATHER

ACTED_IN ACTED_IN

Person

bornIn = 'New Jersey'chauffeurName = 'John Brown'name = 'Michael Douglas'

ACTED_IN ACTED_IN

Figure 15. Graph

9.2.2. Merge nodes

Merge single node with a label

Merging a single node with a given label.

Query

MERGE (robert:Critic)RETURN robert, labels(robert)

A new node is created because there are no nodes labeled Critic in the database.

Result

+----------------------------+| robert | labels(robert) |+----------------------------+| Node[7]{} | ["Critic"] |+----------------------------+1 rowNodes created: 1Labels added: 1

Merge single node with properties

Merging a single node with properties where not all properties match any existing node.

Query

MERGE (charlie { name:'Charlie Sheen', age:10 })RETURN charlie

A new node with the name 'Charlie Sheen' will be created since not all properties matched the existing'Charlie Sheen' node.

Result

+--------------------------------------+| charlie |+--------------------------------------+| Node[7]{name:"Charlie Sheen",age:10} |+--------------------------------------+1 rowNodes created: 1Properties set: 2

106

Merge single node specifying both label and property

Merging a single node with both label and property matching an existing node.

Query

MERGE (michael:Person { name:'Michael Douglas' })RETURN michael.name, michael.bornIn

'Michael Douglas' will be matched and the name and bornIn properties returned.

Result

+------------------------------------+| michael.name | michael.bornIn |+------------------------------------+| "Michael Douglas" | "New Jersey" |+------------------------------------+1 row

Merge single node derived from an existing node property

For some property 'p' in each bound node in a set of nodes, a single new node is created for eachunique value for 'p'.

Query

MATCH (person:Person)MERGE (city:City { name: person.bornIn })RETURN person.name, person.bornIn, city

Three nodes labeled City are created, each of which contains a 'name' property with the value of 'NewYork', 'Ohio', and 'New Jersey', respectively. Note that even though the MATCH clause results in threebound nodes having the value 'New York' for the 'bornIn' property, only a single 'New York' node (i.e. aCity node with a name of 'New York') is created. As the 'New York' node is not matched for the firstbound node, it is created. However, the newly-created 'New York' node is matched and bound for thesecond and third bound nodes.

Result

+----------------------------------------------------------------+| person.name | person.bornIn | city |+----------------------------------------------------------------+| "Rob Reiner" | "New York" | Node[7]{name:"New York"} || "Oliver Stone" | "New York" | Node[7]{name:"New York"} || "Charlie Sheen" | "New York" | Node[7]{name:"New York"} || "Michael Douglas" | "New Jersey" | Node[8]{name:"New Jersey"} || "Martin Sheen" | "Ohio" | Node[9]{name:"Ohio"} |+----------------------------------------------------------------+5 rowsNodes created: 3Properties set: 3Labels added: 3

9.2.3. Use ON CREATE and ON MATCH

Merge with ON CREATE

Merge a node and set properties if the node needs to be created.

107

Query

MERGE (keanu:Person { name:'Keanu Reeves' })ON CREATE SET keanu.created = timestamp()RETURN keanu.name, keanu.created

The query creates the 'keanu' node and sets a timestamp on creation time.

Result

+--------------------------------+| keanu.name | keanu.created |+--------------------------------+| "Keanu Reeves" | 1462500491262 |+--------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

Merge with ON MATCH

Merging nodes and setting properties on found nodes.

Query

MERGE (person:Person)ON MATCH SET person.found = TRUE RETURN person.name, person.found

The query finds all the Person nodes, sets a property on them, and returns them.

Result

+----------------------------------+| person.name | person.found |+----------------------------------+| "Rob Reiner" | true || "Oliver Stone" | true || "Charlie Sheen" | true || "Michael Douglas" | true || "Martin Sheen" | true |+----------------------------------+5 rowsProperties set: 5

Merge with ON CREATE and ON MATCH

Merge a node and set properties if the node needs to be created.

Query

MERGE (keanu:Person { name:'Keanu Reeves' })ON CREATE SET keanu.created = timestamp()ON MATCH SET keanu.lastSeen = timestamp()RETURN keanu.name, keanu.created, keanu.lastSeen

The query creates the 'keanu' node, and sets a timestamp on creation time. If 'keanu' had alreadyexisted, a different property would have been set.

108

Result

+-------------------------------------------------+| keanu.name | keanu.created | keanu.lastSeen |+-------------------------------------------------+| "Keanu Reeves" | 1462500494945 | <null> |+-------------------------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

Merge with ON MATCH setting multiple properties

If multiple properties should be set, simply separate them with commas.

Query

MERGE (person:Person)ON MATCH SET person.found = TRUE , person.lastAccessed = timestamp()RETURN person.name, person.found, person.lastAccessed

Result

+--------------------------------------------------------+| person.name | person.found | person.lastAccessed |+--------------------------------------------------------+| "Rob Reiner" | true | 1462500493713 || "Oliver Stone" | true | 1462500493713 || "Charlie Sheen" | true | 1462500493713 || "Michael Douglas" | true | 1462500493713 || "Martin Sheen" | true | 1462500493713 |+--------------------------------------------------------+5 rowsProperties set: 10

9.2.4. Merge relationships

Merge on a relationship

MERGE can be used to match or create a relationship.

Query

MATCH (charlie:Person { name:'Charlie Sheen' }),(wallStreet:Movie { title:'Wall Street' })MERGE (charlie)-[r:ACTED_IN]->(wallStreet)RETURN charlie.name, type(r), wallStreet.title

'Charlie Sheen' had already been marked as acting in 'Wall Street', so the existing relationship is foundand returned. Note that in order to match or create a relationship when using MERGE, at least onebound node must be specified, which is done via the MATCH clause in the above example.

Result

+-------------------------------------------------+| charlie.name | type(r) | wallStreet.title |+-------------------------------------------------+| "Charlie Sheen" | "ACTED_IN" | "Wall Street" |+-------------------------------------------------+1 row

109

Merge on multiple relationships

When MERGE is used on a whole pattern, either everything matches, or everything is created.

Query

MATCH (oliver:Person { name:'Oliver Stone' }),(reiner:Person { name:'Rob Reiner' })MERGE (oliver)-[:DIRECTED]->(movie:Movie)<-[:ACTED_IN]-(reiner)RETURN movie

In our example graph, 'Oliver Stone' and 'Rob Reiner' have never worked together. When we try toMERGE a movie between them, Neo4j will not use any of the existing movies already connected toeither person. Instead, a new 'movie' node is created.

Result

+-----------+| movie |+-----------+| Node[7]{} |+-----------+1 rowNodes created: 1Relationships created: 2Labels added: 1

Merge on an undirected relationship

MERGE can also be used with an undirected relationship. When it needs to create a new one, it willpick a direction.

Query

MATCH (charlie:Person { name:'Charlie Sheen' }),(oliver:Person { name:'Oliver Stone' })MERGE (charlie)-[r:KNOWS]-(oliver)RETURN r

As 'Charlie Sheen' and 'Oliver Stone' do not know each other, this MERGE query will create a :KNOWSrelationship between them. The direction of the created relationship is arbitrary.

Result

+-------------+| r |+-------------+| :KNOWS[8]{} |+-------------+1 rowRelationships created: 1

Merge on a relationship between two existing nodes

MERGE can be used in conjunction with preceding MATCH and MERGE clauses to create a relationshipbetween two bound nodes 'm' and 'n', where 'm' is returned by MATCH and 'n' is created or matchedby the earlier MERGE.

Query

MATCH (person:Person)MERGE (city:City { name: person.bornIn })MERGE (person)-[r:BORN_IN]->(city)RETURN person.name, person.bornIn, city

110

This builds on the example from Merge single node derived from an existing node property. Thesecond MERGE creates a BORN_IN relationship between each person and a city corresponding to thevalue of the person’s 'bornIn' property. 'Charlie Sheen', 'Rob Reiner' and 'Oliver Stone' all have aBORN_IN relationship to the 'same' City node ('New York').

Result

+----------------------------------------------------------------+| person.name | person.bornIn | city |+----------------------------------------------------------------+| "Rob Reiner" | "New York" | Node[7]{name:"New York"} || "Oliver Stone" | "New York" | Node[7]{name:"New York"} || "Charlie Sheen" | "New York" | Node[7]{name:"New York"} || "Michael Douglas" | "New Jersey" | Node[8]{name:"New Jersey"} || "Martin Sheen" | "Ohio" | Node[9]{name:"Ohio"} |+----------------------------------------------------------------+5 rowsNodes created: 3Relationships created: 5Properties set: 3Labels added: 3

Merge on a relationship between an existing node and a merged node derivedfrom a node property

MERGE can be used to simultaneously create both a new node 'n' and a relationship between a boundnode 'm' and 'n'.

Query

MATCH (person:Person)MERGE (person)-[r:HAS_CHAUFFEUR]->(chauffeur:Chauffeur { name: person.chauffeurName })RETURN person.name, person.chauffeurName, chauffeur

As MERGE found no matches — in our example graph, there are no nodes labeled with Chauffeur andno HAS_CHAUFFEUR relationships — MERGE creates five nodes labeled with Chauffeur, each of whichcontains a 'name' property whose value corresponds to each matched Person node’s 'chauffeurName'property value. MERGE also creates a HAS_CHAUFFEUR relationship between each Person node andthe newly-created corresponding Chauffeur node. As 'Charlie Sheen' and 'Michael Douglas' both havea chauffeur with the same name — 'John Brown' — a new node is created in each case, resulting in'two' Chauffeur nodes having a 'name' of 'John Brown', correctly denoting the fact that even thoughthe 'name' property may be identical, these are two separate people. This is in contrast to the exampleshown above in Merge on a relationship between two existing nodes, where we used the first MERGEto bind the City nodes to prevent them from being recreated (and thus duplicated) in the secondMERGE.

Result

+------------------------------------------------------------------------+| person.name | person.chauffeurName | chauffeur |+------------------------------------------------------------------------+| "Rob Reiner" | "Ted Green" | Node[7]{name:"Ted Green"} || "Oliver Stone" | "Bill White" | Node[8]{name:"Bill White"} || "Charlie Sheen" | "John Brown" | Node[9]{name:"John Brown"} || "Michael Douglas" | "John Brown" | Node[10]{name:"John Brown"} || "Martin Sheen" | "Bob Brown" | Node[11]{name:"Bob Brown"} |+------------------------------------------------------------------------+5 rowsNodes created: 5Relationships created: 5Properties set: 5Labels added: 5

111

9.2.5. Using unique constraints with MERGE

Cypher prevents getting conflicting results from MERGE when using patterns that involve uniquenessconstrains. In this case, there must be at most one node that matches that pattern.

For example, given two uniqueness constraints on :Person(id) and :Person(ssn): then a query such asMERGE (n:Person {id: 12, ssn: 437}) will fail, if there are two different nodes (one with id 12 and onewith ssn 437) or if there is only one node with only one of the properties. In other words, there mustbe exactly one node that matches the pattern, or no matching nodes.

Note that the following examples assume the existence of uniqueness constraints that have beencreated using:

CREATE CONSTRAINT ON (n:Person) ASSERT n.name IS UNIQUE;CREATE CONSTRAINT ON (n:Person) ASSERT n.role IS UNIQUE;

Merge using unique constraints creates a new node if no node is found

Merge using unique constraints creates a new node if no node is found.

Query

MERGE (laurence:Person { name: 'Laurence Fishburne' })RETURN laurence.name

The query creates the 'laurence' node. If 'laurence' had already existed, MERGE would just match theexisting node.

Result

+----------------------+| laurence.name |+----------------------+| "Laurence Fishburne" |+----------------------+1 rowNodes created: 1Properties set: 1Labels added: 1

Merge using unique constraints matches an existing node

Merge using unique constraints matches an existing node.

Query

MERGE (oliver:Person { name:'Oliver Stone' })RETURN oliver.name, oliver.bornIn

The 'oliver' node already exists, so MERGE just matches it.

Result

+--------------------------------+| oliver.name | oliver.bornIn |+--------------------------------+| "Oliver Stone" | "New York" |+--------------------------------+1 row

112

Merge with unique constraints and partial matches

Merge using unique constraints fails when finding partial matches.

Query

MERGE (michael:Person { name:'Michael Douglas', role:'Gordon Gekko' })RETURN michael

While there is a matching unique 'michael' node with the name 'Michael Douglas', there is no uniquenode with the role of 'Gordon Gekko' and MERGE fails to match.

Error message

Merge did not find a matching node and can not create a new node due to conflictswith both existing and missing unique nodes. The conflicting constraints are on::Person.name and :Person.role

Merge with unique constraints and conflicting matches

Merge using unique constraints fails when finding conflicting matches.

Query

MERGE (oliver:Person { name:'Oliver Stone', role:'Gordon Gekko' })RETURN oliver

While there is a matching unique 'oliver' node with the name 'Oliver Stone', there is also anotherunique node with the role of 'Gordon Gekko' and MERGE fails to match.

Error message

Merge did not find a matching node and can not create a new node due to conflictswith both existing and missing unique nodes. The conflicting constraints are on::Person.name and :Person.role

9.2.6. Using map parameters with MERGE

MERGE does not support map parameters like for example CREATE does. To use map parameters withMERGE, it is necessary to explicitly use the expected properties, like in the following example. Formore information on parameters, see Parameters.

Parameters

{ "param" : { "name" : "Keanu Reeves", "role" : "Neo" }}

Query

MERGE (person:Person { name: { param }.name, role: { param }.role })RETURN person.name, person.role

113

Result

+------------------------------+| person.name | person.role |+------------------------------+| "Keanu Reeves" | "Neo" |+------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

9.3. Set

The SET clause is used to update labels on nodes and properties on nodes and relationships.

SET can also be used with maps from parameters to set properties.

Setting labels on a node is an idempotent operations — if you try to set a label on anode that already has that label on it, nothing happens. The query statistics will tellyou if something needed to be done or not.

The examples use this graph as a starting point:

Swedish

name = 'Andres'age = 36hungry = true

name = 'Peter'age = 34

KNOWS

name = 'Emil'

KNOWS

name = 'Stefan'

KNOWS

9.3.1. Set a property

To set a property on a node or relationship, use SET.

Query

MATCH (n { name: 'Andres' })SET n.surname = 'Taylor'RETURN n

The newly changed node is returned by the query.

Result

+------------------------------------------------------------+| n |+------------------------------------------------------------+| Node[0]{surname:"Taylor",name:"Andres",age:36,hungry:true} |+------------------------------------------------------------+1 rowProperties set: 1

114

9.3.2. Remove a property

Normally you remove a property by using <<query-remove,REMOVE>>, but it’s sometimes handy to doit using the SET command. One example is if the property comes from a parameter.

Query

MATCH (n { name: 'Andres' })SET n.name = NULL RETURN n

The node is returned by the query, and the name property is now missing.

Result

+-----------------------------+| n |+-----------------------------+| Node[0]{hungry:true,age:36} |+-----------------------------+1 rowProperties set: 1

9.3.3. Copying properties between nodes and relationships

You can also use SET to copy all properties from one graph element to another. Remember that doingthis will remove all other properties on the receiving graph element.

Query

MATCH (at { name: 'Andres' }),(pn { name: 'Peter' })SET at = pnRETURN at, pn

The Andres node has had all it’s properties replaced by the properties in the Peter node.

Result

+-------------------------------------------------------------+| at | pn |+-------------------------------------------------------------+| Node[0]{name:"Peter",age:34} | Node[2]{name:"Peter",age:34} |+-------------------------------------------------------------+1 rowProperties set: 3

9.3.4. Adding properties from maps

When setting properties from a map (literal, paremeter, or graph element), you can use the += form ofSET to only add properties, and not remove any of the existing properties on the graph element.

Query

MATCH (peter { name: 'Peter' })SET peter += { hungry: TRUE , position: 'Entrepreneur' }

115

Result

+-------------------+| No data returned. |+-------------------+Properties set: 2

9.3.5. Set a property using a parameter

Use a parameter to give the value of a property.

Parameters

{ "surname" : "Taylor"}

Query

MATCH (n { name: 'Andres' })SET n.surname = { surname }RETURN n

The Andres node has got an surname added.

Result

+------------------------------------------------------------+| n |+------------------------------------------------------------+| Node[0]{surname:"Taylor",name:"Andres",age:36,hungry:true} |+------------------------------------------------------------+1 rowProperties set: 1

9.3.6. Set all properties using a parameter

This will replace all existing properties on the node with the new set provided by the parameter.

Parameters


Query

MATCH (n { name: 'Andres' })SET n = { props }RETURN n

The Andres node has had all it’s properties replaced by the properties in the props parameter.

116

Result

+---------------------------------------------+| n |+---------------------------------------------+| Node[0]{name:"Andres",position:"Developer"} |+---------------------------------------------+1 rowProperties set: 4

9.3.7. Set multiple properties using one SET clause

If you want to set multiple properties in one go, simply separate them with a comma.

Query

MATCH (n { name: 'Andres' })SET n.position = 'Developer', n.surname = 'Taylor'

Result


9.3.8. Set a label on a node

To set a label on a node, use SET.

Query

MATCH (n { name: 'Stefan' })SET n :GermanRETURN n

The newly labeled node is returned by the query.

Result

+------------------------+| n |+------------------------+| Node[3]{name:"Stefan"} |+------------------------+1 rowLabels added: 1

9.3.9. Set multiple labels on a node

To set multiple labels on a node, use SET and separate the different labels using :.

Query

MATCH (n { name: 'Emil' })SET n :Swedish:BossmanRETURN n

The newly labeled node is returned by the query.

117

Result

+----------------------+| n |+----------------------+| Node[1]{name:"Emil"} |+----------------------+1 rowLabels added: 2

9.4. Delete

The DELETE clause is used to delete graph elements — nodes, relationships or paths.

For removing properties and labels, see Remove. Remember that you can not delete a node withoutalso deleting relationships that start or end on said node. Either explicitly delete the relationships, oruse DETACH DELETE.

The examples start out with the following database:

name = 'Andres'age = 36


KNOWS

name = 'Tobias'age = 25

KNOWS

9.4.1. Delete single node

To delete a node, use the DELETE clause.

Query

MATCH (n:Useless)DELETE n

Result

+-------------------+| No data returned. |+-------------------+Nodes deleted: 1

9.4.2. Delete all nodes and relationships

This query isn’t for deleting large amounts of data, but is nice when playing around with smallexample data sets.

Query

MATCH (n)DETACH DELETE n

118

Result

+-------------------+| No data returned. |+-------------------+Nodes deleted: 3Relationships deleted: 2

9.4.3. Delete a node with all its relationships

When you want to delete a node and any relationship going to or from it, use DETACH DELETE.

Query

MATCH (n { name:'Andres' })DETACH DELETE n

Result

+-------------------+| No data returned. |+-------------------+Nodes deleted: 1Relationships deleted: 2

9.5. Remove

The REMOVE clause is used to remove properties and labels from graph elements.

For deleting nodes and relationships, see Delete.

Removing labels from a node is an idempotent operation: If you try to remove alabel from a node that does not have that label on it, nothing happens. The querystatistics will tell you if something needed to be done or not.

The examples start out with the following database:

Swedish

name = 'Andres'age = 36

Swedish, German


KNOWS

Swedish

name = 'Tobias'age = 25

KNOWS

9.5.1. Remove a property

Neo4j doesn’t allow storing null in properties. Instead, if no value exists, the property is just not there.So, to remove a property value on a node or a relationship, is also done with REMOVE.

Query

MATCH (andres { name: 'Andres' })REMOVE andres.ageRETURN andres

119

The node is returned, and no property age exists on it.

Result

+------------------------+| andres |+------------------------+| Node[0]{name:"Andres"} |+------------------------+1 rowProperties set: 1

9.5.2. Remove a label from a node

To remove labels, you use REMOVE.

Query

MATCH (n { name: 'Peter' })REMOVE n:GermanRETURN n

Result

+------------------------------+| n |+------------------------------+| Node[2]{name:"Peter",age:34} |+------------------------------+1 rowLabels removed: 1

9.5.3. Removing multiple labels

To remove multiple labels, you use REMOVE.

Query

MATCH (n { name: 'Peter' })REMOVE n:German:SwedishRETURN n

Result

+------------------------------+| n |+------------------------------+| Node[2]{name:"Peter",age:34} |+------------------------------+1 rowLabels removed: 2

9.6. Foreach

The FOREACH clause is used to update data within a list, whether components of a path, or

result of aggregation.

Lists and paths are key concepts in Cypher. To use them for updating data, you can use the FOREACHconstruct. It allows you to do updating commands on elements in a path, or a list created by

120

aggregation.

The variable context inside of the foreach parenthesis is separate from the one outside it. This meansthat if you CREATE a node variable inside of a FOREACH, you will not be able to use it outside of theforeach statement, unless you match to find it.

Inside of the FOREACH parentheses, you can do any of the updating commands — CREATE, CREATEUNIQUE, MERGE, DELETE, and FOREACH.

If you want to execute an additional MATCH for each element in a list then UNWIND (see Unwind) wouldbe a more appropriate command.

name = 'A'

name = 'B'

KNOWS

name = 'C'

KNOWS

name = 'D'

KNOWS

Figure 16. Data for the examples

9.6.1. Mark all nodes along a path

This query will set the property marked to true on all nodes along a path.

Query

MATCH p =(begin)-[*]->(END )WHERE begin.name='A' AND END .name='D'FOREACH (n IN nodes(p)| SET n.marked = TRUE )

Nothing is returned from this query, but four properties are set.

Result


9.7. Create Unique

The CREATE UNIQUE clause is a mix of MATCH and CREATE — it will match what it can, and

create what is missing.

121

9.7.1. Introduction

<<query-merge,MERGE>> might be what you want to use instead of CREATEUNIQUE. Note however, that MERGE doesn’t give as strong guarantees forrelationships being unique.

CREATE UNIQUE is in the middle of MATCH and CREATE — it will match what it can, and create what ismissing. CREATE UNIQUE will always make the least change possible to the graph — if it can use partsof the existing graph, it will.

Another difference to MATCH is that CREATE UNIQUE assumes the pattern to be unique. If multiplematching subgraphs are found an error will be generated.

In the CREATE UNIQUE clause, patterns are used a lot. Read Patterns for anintroduction.

The examples start out with the following data set:

name = 'A'

name = 'C'

KNOWS

name = 'B'

name = 'root'

X

X

X

9.7.2. Create unique nodes

Create node if missing

If the pattern described needs a node, and it can’t be matched, a new node will be created.

Query

MATCH (root { name: 'root' })CREATE UNIQUE (root)-[:LOVES]-(someone)RETURN someone

The root node doesn’t have any LOVES relationships, and so a node is created, and also a relationshipto that node.

Result

+-----------+| someone |+-----------+| Node[4]{} |+-----------+1 rowNodes created: 1Relationships created: 1

122

Create nodes with values

The pattern described can also contain values on the node. These are given using the followingsyntax: prop : <expression>.

Query

MATCH (root { name: 'root' })CREATE UNIQUE (root)-[:X]-(leaf { name:'D' })RETURN leaf

No node connected with the root node has the name D, and so a new node is created to match thepattern.

Result

+-------------------+| leaf |+-------------------+| Node[4]{name:"D"} |+-------------------+1 rowNodes created: 1Relationships created: 1Properties set: 1

Create labeled node if missing

If the pattern described needs a labeled node and there is none with the given labels, Cypher willcreate a new one.

Query

MATCH (a { name: 'A' })CREATE UNIQUE (a)-[:KNOWS]-(c:blue)RETURN c

The A node is connected in a KNOWS relationship to the c node, but since C doesn’t have the :blue label,a new node labeled as :blue is created along with a KNOWS relationship from A to it.

Result

+-----------+| c |+-----------+| Node[4]{} |+-----------+1 rowNodes created: 1Relationships created: 1Labels added: 1

9.7.3. Create unique relationships

Create relationship if it is missing

CREATE UNIQUE is used to describe the pattern that should be found or created.

123

Query

MATCH (lft { name: 'A' }),(rgt)WHERE rgt.name IN ['B', 'C']CREATE UNIQUE (lft)-[r:KNOWS]->(rgt)RETURN r

The left node is matched agains the two right nodes. One relationship already exists and can bematched, and the other relationship is created before it is returned.

Result

+-------------+| r |+-------------+| :KNOWS[4]{} || :KNOWS[3]{} |+-------------+2 rowsRelationships created: 1

Create relationship with values

Relationships to be created can also be matched on values.

Query

MATCH (root { name: 'root' })CREATE UNIQUE (root)-[r:X { since:'forever' }]-()RETURN r

In this example, we want the relationship to have a value, and since no such relationship can befound, a new node and relationship are created. Note that since we are not interested in the creatednode, we don’t name it.

Result

+------------------------+| r |+------------------------+| :X[4]{since:"forever"} |+------------------------+1 rowNodes created: 1Relationships created: 1Properties set: 1

9.7.4. Describe complex pattern

The pattern described by CREATE UNIQUE can be separated by commas, just like in MATCH andCREATE.

Query

MATCH (root { name: 'root' })CREATE UNIQUE (root)-[:FOO]->(x),(root)-[:BAR]->(x)RETURN x

This example pattern uses two paths, separated by a comma.

124

Result

+-----------+| x |+-----------+| Node[4]{} |+-----------+1 rowNodes created: 1Relationships created: 2

9.8. Importing CSV files with Cypher

This tutorial will show you how to import data from CSV files using LOAD CSV.

In this example, we’re given three CSV files: a list of persons, a list of movies, and a list of which rolewas played by some of these persons in each movie.

CSV files can be stored on the database server and are then accessible using a file:// URL.Alternatively, LOAD CSV also supports accessing CSV files via HTTPS, HTTP, and FTP. LOAD CSV willfollow HTTP redirects but for security reasons it will not follow redirects that changes the protocol, forexample if the redirect is going from HTTPS to HTTP.

For more details, see Load CSV.

Using the following Cypher queries, we’ll create a node for each person, a node for each movie and arelationship between the two with a property denoting the role. We’re also keeping track of thecountry in which each movie was made.

Let’s start with importing the persons:

LOAD CSV WITH HEADERS FROM "http://neo4j.com/docs/3.0.1/csv/import/persons.csv" AS csvLineCREATE (p:Person { id: toInt(csvLine.id), name: csvLine.name })

The CSV file we’re using looks like this:

persons.csv

id,name1,Charlie Sheen2,Oliver Stone3,Michael Douglas4,Martin Sheen5,Morgan Freeman

Now, let’s import the movies. This time, we’re also creating a relationship to the country in which themovie was made. If you are storing your data in a SQL database, this is the one-to-many relationshiptype.

We’re using MERGE to create nodes that represent countries. Using MERGE avoids creating duplicatecountry nodes in the case where multiple movies have been made in the same country.

When using MERGE or MATCH with LOAD CSV we need to make sure we have anindex (see Indexes) or a unique constraint (see Constraints) on the property we’remerging. This will ensure the query executes in a performant way.

Before running our query to connect movies and countries we’ll create an index for the name propertyon the Country label to ensure the query runs as fast as it can:

125

CREATE INDEX ON :Country(name)

LOAD CSV WITH HEADERS FROM "http://neo4j.com/docs/3.0.1/csv/import/movies.csv" AS csvLineMERGE (country:Country { name: csvLine.country })CREATE (movie:Movie { id: toInt(csvLine.id), title: csvLine.title, year:toInt(csvLine.year)})CREATE (movie)-[:MADE_IN]->(country)

movies.csv

id,title,country,year1,Wall Street,USA,19872,The American President,USA,19953,The Shawshank Redemption,USA,1994

Lastly, we create the relationships between the persons and the movies. Since the relationship is amany to many relationship, one actor can participate in many movies, and one movie has many actorsin it. We have this data in a separate file.

We’ll index the id property on Person and Movie nodes. The id property is a temporary property usedto look up the appropriate nodes for a relationship when importing the third file. By indexing the idproperty, node lookup (e.g. by MATCH) will be much faster. Since we expect the ids to be unique ineach set, we’ll create a unique constraint. This protects us from invalid data since constraint creationwill fail if there are multiple nodes with the same id property. Creating a unique constraint also createsa unique index (which is faster than a regular index).

CREATE CONSTRAINT ON (person:Person) ASSERT person.id IS UNIQUE

CREATE CONSTRAINT ON (movie:Movie) ASSERT movie.id IS UNIQUE

Now importing the relationships is a matter of finding the nodes and then creating relationshipsbetween them.

For this query we’ll use USING PERIODIC COMMIT (see Using Periodic Commit) which is helpful forqueries that operate on large CSV files. This hint tells Neo4j that the query might build up inordinateamounts of transaction state, and so needs to be periodically committed. In this case we also set thelimit to 500 rows per commit.

USING PERIODIC COMMIT 500LOAD CSV WITH HEADERS FROM "http://neo4j.com/docs/3.0.1/csv/import/roles.csv" AS csvLineMATCH (person:Person { id: toInt(csvLine.personId)}),(movie:Movie { id: toInt(csvLine.movieId)})CREATE (person)-[:PLAYED { role: csvLine.role }]->(movie)

roles.csv

personId,movieId,role1,1,Bud Fox4,1,Carl Fox3,1,Gordon Gekko4,2,A.J. MacInerney3,2,President Andrew Shepherd5,3,Ellis Boyd 'Red' Redding

Finally, as the id property was only necessary to import the relationships, we can drop the constraintsand the id property from all movie and person nodes.

DROP CONSTRAINT ON (person:Person) ASSERT person.id IS UNIQUE

126

DROP CONSTRAINT ON (movie:Movie) ASSERT movie.id IS UNIQUE

MATCH (n)WHERE n:Person OR n:MovieREMOVE n.id

9.9. Using Periodic Commit

See Importing CSV files with Cypher on how to import data from CSV files.

Importing large amounts of data using LOAD CSV with a single Cypher query may fail due to memoryconstraints. This will manifest itself as an OutOfMemoryError.

For this situation only, Cypher provides the global USING PERIODIC COMMIT query hint for updatingqueries using LOAD CSV. You can optionally set the limit for the number of rows per commit like so:USING PERIODIC COMMIT 500.

PERIODIC COMMIT will process the rows until the number of rows reaches a limit. Then the currenttransaction will be committed and replaced with a newly opened transaction. If no limit is set, adefault value will be used.

See Importing large amounts of data in Load CSV for examples of USING PERIODIC COMMIT with andwithout setting the number of rows per commit.

Using periodic commit will prevent running out of memory when importing largeamounts of data. However, it will also break transactional isolation and thus itshould only be used where needed.

127

Chapter 10. FunctionsThis chapter contains information on all functions in Cypher. Note that related information exists inOperators.

Most functions in Cypher will return NULL if an input parameter is NULL.

10.1. Predicates

Predicates are boolean functions that return true or false for a given set of input. They are mostcommonly used to filter out subgraphs in the WHERE part of a query.

See also Comparison operators.

bar, foo

name = 'Alice'age = 38eyes = 'brown'

name = 'Charlie'age = 53eyes = 'green'

KNOWS

name = 'Bob'age = 25eyes = 'blue'

KNOWS

name = 'Daniel'age = 54eyes = 'brown'

KNOWS

Spouse

array = ['one', 'two', 'three']name = 'Eskil'age = 41eyes = 'blue'

MARRIEDKNOWS

Figure 17. Graph

10.1.1. all()

Tests whether a predicate holds for all elements of this list.

Syntax: all(variable IN list WHERE predicate)

Arguments:

• list: An expression that returns a list

• variable: This is the variable that can be used from the predicate.

• predicate: A predicate that is tested against all items in the list.

Query

MATCH p=(a)-[*1..3]->(b)WHERE a.name='Alice' AND b.name='Daniel' AND ALL (x IN nodes(p) WHERE x.age > 30)RETURN p

All nodes in the returned paths will have an age property of at least 30.

128

Result

+------------------------------------------------------------------------------------------------------------------------------------------------------------+| p|+------------------------------------------------------------------------------------------------------------------------------------------------------------+|[Node[0]{name:"Alice",age:38,eyes:"brown"},:KNOWS[1]{},Node[2]{name:"Charlie",age:53,eyes:"green"},:KNOWS[3]{},Node[3]{name:"Daniel",age:54,eyes:"brown"}] |+------------------------------------------------------------------------------------------------------------------------------------------------------------+1 row

10.1.2. any()

Tests whether a predicate holds for at least one element in the list.

Syntax: any(variable IN list WHERE predicate)

Arguments:




Query

MATCH (a)WHERE a.name='Eskil' AND ANY (x IN a.array WHERE x = "one")RETURN a

All nodes in the returned paths has at least one one value set in the array property named array.

Result

+----------------------------------------------------------------------+| a |+----------------------------------------------------------------------+| Node[4]{array:["one","two","three"],name:"Eskil",age:41,eyes:"blue"} |+----------------------------------------------------------------------+1 row

10.1.3. none()

Returns true if the predicate holds for no element in the list.

Syntax: none(variable in list WHERE predicate)

Arguments:




129

Query

MATCH p=(n)-[*1..3]->(b)WHERE n.name='Alice' AND NONE (x IN nodes(p) WHERE x.age = 25)RETURN p

No nodes in the returned paths has a age property set to 25.

Result

+------------------------------------------------------------------------------------------------------------------------------------------------------------+| p|+------------------------------------------------------------------------------------------------------------------------------------------------------------+| [Node[0]{name:"Alice",age:38,eyes:"brown"},:KNOWS[1]{},Node[2]{name:"Charlie",age:53,eyes:"green"}]||[Node[0]{name:"Alice",age:38,eyes:"brown"},:KNOWS[1]{},Node[2]{name:"Charlie",age:53,eyes:"green"},:KNOWS[3]{},Node[3]{name:"Daniel",age:54,eyes:"brown"}] |+------------------------------------------------------------------------------------------------------------------------------------------------------------+2 rows

10.1.4. single()

Returns true if the predicate holds for exactly one of the elements in the list.

Syntax: single(variable in list WHERE predicate)

Arguments:




Query

MATCH p=(n)-->(b)WHERE n.name='Alice' AND SINGLE (var IN nodes(p) WHERE var.eyes = "blue")RETURN p

Exactly one node in every returned path will have the eyes property set to "blue".

Result

+------------------------------------------------------------------------------------------------+| p |+------------------------------------------------------------------------------------------------+| [Node[0]{name:"Alice",age:38,eyes:"brown"},:KNOWS[0]{},Node[1]{name:"Bob",age:25,eyes:"blue"}] |+------------------------------------------------------------------------------------------------+1 row

10.1.5. exists()

Returns true if a match for the pattern exists in the graph, or the property exists in the node,relationship or map.

Syntax: exists( pattern-or-property )

130

Arguments:

• pattern-or-property: A pattern or a property (in the form 'variable.prop').

Query

MATCH (n)WHERE EXISTS(n.name)RETURN n.name AS name, EXISTS((n)-[:MARRIED]->()) AS is_married

This query returns all the nodes with a name property along with a boolean true/false indicating ifthey are married.

Result

+------------------------+| name | is_married |+------------------------+| "Alice" | false || "Bob" | true || "Charlie" | false || "Daniel" | false || "Eskil" | false |+------------------------+5 rows

10.2. Scalar functions

Scalar functions return a single value.

The length() and size() functions are quite similar, and so it is important to takenote of the difference. Due to backwards compatibility, length() currently works onfour types: strings, paths, lists and pattern expressions. However, for clarity it isrecommended to only use length() on strings and paths, and use the new size()function on lists and pattern expressions. length() on those types may bedeprecated in future.

bar, foo



KNOWS


KNOWS


KNOWS

Spouse


MARRIEDKNOWS

Figure 18. Graph

10.2.1. size()

To return or filter on the size of a list, use the size() function.

Syntax: size( list )

131

Arguments:


Query

RETURN size(['Alice', 'Bob']) AS col

The number of items in the list is returned by the query.

Result

+-----+| col |+-----+| 2 |+-----+1 row

10.2.2. Size of pattern expression

This is the same size() method described before, but instead of passing in a list directly, you providea pattern expression that can be used in a match query to provide a new set of results. The size of theresult is calculated, not the length of the expression itself.

Syntax: size( pattern expression )

Arguments:

• pattern expression: A pattern expression that returns a list

Query

MATCH (a)WHERE a.name='Alice'RETURN size((a)-->()-->()) AS fof

The number of sub-graphs matching the pattern expression is returned by the query.

Result

+-----+| fof |+-----+| 3 |+-----+1 row

10.2.3. length()

To return or filter on the length of a path, use the LENGTH() function.

Syntax: length( path )

Arguments:

• path: An expression that returns a path

132

Query

MATCH p=(a)-->(b)-->(c)WHERE a.name='Alice'RETURN length(p)

The length of the path p is returned by the query.

Result

+-----------+| length(p) |+-----------+| 2 || 2 || 2 |+-----------+3 rows

10.2.4. Length of string

To return or filter on the length of a string, use the LENGTH() function.

Syntax: length( string )

Arguments:

• string: An expression that returns a string

Query

MATCH (a)WHERE length(a.name)> 6RETURN length(a.name)

The length of the name Charlie is returned by the query.

Result

+----------------+| length(a.name) |+----------------+| 7 |+----------------+1 row

10.2.5. type()

Returns a string representation of the relationship type.

Syntax: type( relationship )

Arguments:

• relationship: A relationship.

Query

MATCH (n)-[r]->()WHERE n.name='Alice'RETURN type(r)

133

The relationship type of r is returned by the query.

Result

+---------+| type(r) |+---------+| "KNOWS" || "KNOWS" |+---------+2 rows

10.2.6. id()

Returns the id of the relationship or node.

Syntax: id( property-container )

Arguments:

• property-container: A node or a relationship.

Query

MATCH (a)RETURN id(a)

This returns the node id for three nodes.

Result

+-------+| id(a) |+-------+| 0 || 1 || 2 || 3 || 4 |+-------+5 rows

10.2.7. coalesce()

Returns the first non-NULL value in the list of expressions passed to it. In case all arguments are NULL,NULL will be returned.

Syntax: coalesce( expression [, expression]* )

Arguments:

• expression: The expression that might return NULL.

Query

MATCH (a)WHERE a.name='Alice'RETURN coalesce(a.hairColor, a.eyes)

134

Result

+-------------------------------+| coalesce(a.hairColor, a.eyes) |+-------------------------------+| "brown" |+-------------------------------+1 row

10.2.8. head()

head() returns the first element in a list.

Syntax: head( expression )

Arguments:

• expression: This expression should return a list of some kind.

Query

MATCH (a)WHERE a.name='Eskil'RETURN a.array, head(a.array)

The first node in the path is returned.

Result

+---------------------------------------+| a.array | head(a.array) |+---------------------------------------+| ["one","two","three"] | "one" |+---------------------------------------+1 row

10.2.9. last()

last() returns the last element in a list.

Syntax: last( expression )

Arguments:


Query

MATCH (a)WHERE a.name='Eskil'RETURN a.array, last(a.array)

The last node in the path is returned.

135

Result

+---------------------------------------+| a.array | last(a.array) |+---------------------------------------+| ["one","two","three"] | "three" |+---------------------------------------+1 row

10.2.10. timestamp()

timestamp() returns the difference, measured in milliseconds, between the current time and midnight,January 1, 1970 UTC. It will return the same value during the whole one query, even if the query is along running one.

Syntax: timestamp()

Arguments:

Query

RETURN timestamp()

The time in milliseconds is returned.

Result

+---------------+| timestamp() |+---------------+| 1462500541123 |+---------------+1 row

10.2.11. startNode()

startNode() returns the starting node of a relationship

Syntax: startNode( relationship )

Arguments:

• relationship: An expression that returns a relationship

Query

MATCH (x:foo)-[r]-()RETURN startNode(r)

Result

+-------------------------------------------+| startNode(r) |+-------------------------------------------+| Node[0]{name:"Alice",age:38,eyes:"brown"} || Node[0]{name:"Alice",age:38,eyes:"brown"} |+-------------------------------------------+2 rows

136

10.2.12. endNode()

endNode() returns the end node of a relationship

Syntax: endNode( relationship )

Arguments:

• relationship: An expression that returns a relationship

Query

MATCH (x:foo)-[r]-()RETURN endNode(r)

Result

+---------------------------------------------+| endNode(r) |+---------------------------------------------+| Node[2]{name:"Charlie",age:53,eyes:"green"} || Node[1]{name:"Bob",age:25,eyes:"blue"} |+---------------------------------------------+2 rows

10.2.13. properties()

properties() converts the arguments to a map of its properties. If the argument is a node or arelationship, the returned map is a map of its properties .If the argument is already a map, it isreturned unchanged.

Syntax: properties( expression )

Arguments:

• expression: An expression that returns a node, a relationship, or a map

Query

CREATE (p:Person { name: 'Stefan', city: 'Berlin' })RETURN properties(p)

Result

+--------------------------------------+| properties(p) |+--------------------------------------+| {name -> "Stefan", city -> "Berlin"} |+--------------------------------------+1 rowNodes created: 1Properties set: 2Labels added: 1

10.2.14. toInt()

toInt() converts the argument to an integer. A string is parsed as if it was an integer number. If theparsing fails, NULL will be returned. A floating point number will be cast into an integer.

Syntax: toInt( expression )

137

Arguments:

• expression: An expression that returns anything

Query

RETURN toInt("42"), toInt("not a number")

Result

+-------------------------------------+| toInt("42") | toInt("not a number") |+-------------------------------------+| 42 | <null> |+-------------------------------------+1 row

10.2.15. toFloat()

toFloat() converts the argument to a float. A string is parsed as if it was an floating point number. Ifthe parsing fails, NULL will be returned. An integer will be cast to a floating point number.

Syntax: toFloat( expression )

Arguments:

• expression: An expression that returns anything

Query

RETURN toFloat("11.5"), toFloat("not a number")

Result

+-------------------------------------------+| toFloat("11.5") | toFloat("not a number") |+-------------------------------------------+| 11.5 | <null> |+-------------------------------------------+1 row

10.3. List functions

List functions return lists of things — nodes in a path, and so on.

See also List operators.

138

bar, foo



KNOWS


KNOWS


KNOWS

Spouse


MARRIEDKNOWS

Figure 19. Graph

10.3.1. nodes()

Returns all nodes in a path.

Syntax: nodes( path )

Arguments:

• path: A path.

Query

MATCH p=(a)-->(b)-->(c)WHERE a.name='Alice' AND c.name='Eskil'RETURN nodes(p)

All the nodes in the path p are returned by the example query.

Result

+---------------------------------------------------------------------------------------------------------------------------------------------------------+| nodes(p)|+---------------------------------------------------------------------------------------------------------------------------------------------------------+|[Node[0]{name:"Alice",age:38,eyes:"brown"},Node[1]{name:"Bob",age:25,eyes:"blue"},Node[4]{array:["one","two","three"],name:"Eskil",age:41,eyes:"blue"}] |+---------------------------------------------------------------------------------------------------------------------------------------------------------+1 row

10.3.2. relationships()

Returns all relationships in a path.

Syntax: relationships( path )

Arguments:

• path: A path.

139

Query

MATCH p=(a)-->(b)-->(c)WHERE a.name='Alice' AND c.name='Eskil'RETURN relationships(p)

All the relationships in the path p are returned.

Result

+-----------------------------+| relationships(p) |+-----------------------------+| [:KNOWS[0]{},:MARRIED[4]{}] |+-----------------------------+1 row

10.3.3. labels()

Returns a list of string representations for the labels attached to a node.

Syntax: labels( node )

Arguments:

• node: Any expression that returns a single node

Query

MATCH (a)WHERE a.name='Alice'RETURN labels(a)

The labels of n is returned by the query.

Result

+---------------+| labels(a) |+---------------+| ["bar","foo"] |+---------------+1 row

10.3.4. keys()

Returns a list of string representations for the property names of a node, relationship, or map.

Syntax: keys( property-container )

Arguments:

• property-container: A node, a relationship, or a literal map.

Query

MATCH (a)WHERE a.name='Alice'RETURN keys(a)

140

The name of the properties of n is returned by the query.

Result

+-----------------------+| keys(a) |+-----------------------+| ["name","age","eyes"] |+-----------------------+1 row

10.3.5. extract()

To return a single property, or the value of a function from a list of nodes or relationships, you canuse extract(). It will go through a list, run an expression on every element, and return the results in alist with these values. It works like the map method in functional languages such as Lisp and Scala.

Syntax: extract( variable IN list | expression )

Arguments:


• variable: The closure will have a variable introduced in it’s context. Here you decide which variableto use.

• expression: This expression will run once per value in the list, and produces the result list.

Query

MATCH p=(a)-->(b)-->(c)WHERE a.name='Alice' AND b.name='Bob' AND c.name='Daniel'RETURN extract(n IN nodes(p)| n.age) AS extracted

The age property of all nodes in the path are returned.

Result

+------------+| extracted |+------------+| [38,25,54] |+------------+1 row

10.3.6. filter()

filter() returns all the elements in a list that comply to a predicate.

Syntax: filter(variable IN list WHERE predicate)

Arguments:




141

Query

MATCH (a)WHERE a.name='Eskil'RETURN a.array, filter(x IN a.array WHERE size(x)= 3)

This returns the property named array and a list of values in it, which have size 3.

Result

+----------------------------------------------------------------+| a.array | filter(x in a.array WHERE size(x) = 3) |+----------------------------------------------------------------+| ["one","two","three"] | ["one","two"] |+----------------------------------------------------------------+1 row

10.3.7. tail()

tail() returns all but the first element in a list.

Syntax: tail( expression )

Arguments:


Query

MATCH (a)WHERE a.name='Eskil'RETURN a.array, tail(a.array)

This returns the property named array and all elements of that property except the first one.

Result

+-----------------------------------------+| a.array | tail(a.array) |+-----------------------------------------+| ["one","two","three"] | ["two","three"] |+-----------------------------------------+1 row

10.3.8. range()

range() returns numerical values in a range. The default distance between values in the range is 1. Ther is inclusive in both ends.

Syntax: range( start, end [, step] )

Arguments:

• start: A numerical expression.

• end: A numerical expression.

• step: A numerical expression.

142

Query

RETURN range(0,10), range(2,18,3)

Two lists of numbers in the given ranges are returned.

Result

+---------------------------------------------+| range(0,10) | range(2,18,3) |+---------------------------------------------+| [0,1,2,3,4,5,6,7,8,9,10] | [2,5,8,11,14,17] |+---------------------------------------------+1 row

10.3.9. reduce()

To run an expression against individual elements of a list, and store the result of the expression in anaccumulator, you can use reduce(). It will go through a list, run an expression on every element,storing the partial result in the accumulator. It works like the fold or reduce method in functionallanguages such as Lisp and Scala.

Syntax: reduce( accumulator = initial, variable IN list | expression )

Arguments:

• accumulator: A variable that will hold the result and the partial results as the list is iterated

• initial: An expression that runs once to give a starting value to the accumulator


• variable: The closure will have a variable introduced in it’s context. Here you decide which variableto use.

• expression: This expression will run once per value in the list, and produces the result value.

Query

MATCH p=(a)-->(b)-->(c)WHERE a.name='Alice' AND b.name='Bob' AND c.name='Daniel'RETURN reduce(totalAge = 0, n IN nodes(p)| totalAge + n.age) AS reduction

The age property of all nodes in the path are summed and returned as a single value.

Result

+-----------+| reduction |+-----------+| 117 |+-----------+1 row

10.4. Math functions

These functions all operate on numerical expressions only, and will return an error if used on

any other values. See also Mathematical operators..


143

A


C


KNOWS

B


KNOWS

D


KNOWS

E


MARRIEDKNOWS

Figure 20. Graph

10.4.1. Number functions

abs()

abs() returns the absolute value for a number.

Syntax: abs( expression )

Arguments:

• expression: A numeric expression.

Query

MATCH (a),(e)WHERE a.name = 'Alice' AND e.name = 'Eskil'RETURN a.age, e.age, abs(a.age - e.age)

The absolute value of the age difference is returned.

Table 25. Result

a.age e.age abs(a.age - e.age)

38 41 3

1 row

ceil()

ceil() returns the smallest integer greater than or equal to the argument.

Syntax: ceil( expression )

Arguments:


Query

RETURN ceil(0.1)

144

The ceil of 0.1.

Table 26. Result

ceil(0.1)

1.0

1 row

floor()

floor() returns the greatest integer less than or equal to the expression.

Syntax: floor( expression )

Arguments:


Query

RETURN floor(0.9)

The floor of 0.9 is returned.

Table 27. Result

floor(0.9)

0.0

1 row

round()

round() returns the numerical expression, rounded to the nearest integer.

Syntax: round( expression )

Arguments:

• expression: A numeric expression that represents the angle in radians.

Query

RETURN round(3.141592)

3.0 is returned.

Table 28. Result

round(3.141592)

3.0

1 row

sign()

sign() returns the signum of a number — zero if the expression is zero, -1 for any negative number,

145

and 1 for any positive number.

Syntax: sign( expression )

Arguments:


Query

RETURN sign(-17), sign(0.1)

The signs of -17 and 0.1 are returned.

Table 29. Result

sign(-17) sign(0.1)

-1 1

1 row

rand()

rand() returns a random number in the range from 0 (inclusive) to 1 (exclusive), [0,1). The numbersreturned follow an approximate uniform distribution.

Syntax: rand()

Arguments:

Query

RETURN rand()

A random number is returned.

Table 30. Result

rand()

0.04166747260084902

1 row

10.4.2. Logarithmic functions

log()

log() returns the natural logarithm of the expression.

Syntax: log( expression )

Arguments:


146

Query

RETURN log(27)

The natural logarithm of 27 is returned.

Table 31. Result

log(27)

3.295836866004329

1 row

log10()

log10() returns the common logarithm (base 10) of the expression.

Syntax: log10( expression )

Arguments:


Query

RETURN log10(27)

The common logarithm of 27 is returned.

Table 32. Result

log10(27)

1.4313637641589874

1 row

exp()

exp() returns e^n, where e is the base of the natural logarithm, and n is the value of the argumentexpression.

Syntax: e( expression )

Arguments:


Query

RETURN exp(2)

e to the power of 2 is returned.

Table 33. Result

exp(2)

7.38905609893065

147

exp(2)

1 row

e()

e() returns the base of the natural logarithm, e.

Syntax: e()

Arguments:

Query

RETURN e()

The base of the natural logarithm, e, is returned.

Table 34. Result

e()

2.718281828459045

1 row

sqrt()

sqrt() returns the square root of a number.

Syntax: sqrt( expression )

Arguments:


Query

RETURN sqrt(256)

The square root of 256 is returned.

Table 35. Result

sqrt(256)

16.0

1 row

10.4.3. Trigonometric functions

All trigonometric functions operate on radians, unless otherwise specified.

sin()

sin() returns the sine of the expression.

Syntax: sin( expression )

148

Arguments:


Query

RETURN sin(0.5)

The sine of 0.5 is returned.

Table 36. Result

sin(0.5)

0.479425538604203

1 row

cos()

cos() returns the cosine of the expression.

Syntax: cos( expression )

Arguments:


Query

RETURN cos(0.5)

The cosine of 0.5.

Table 37. Result

cos(0.5)

0.8775825618903728

1 row

tan()

tan() returns the tangent of the expression.

Syntax: tan( expression )

Arguments:


Query

RETURN tan(0.5)

The tangent of 0.5 is returned.

Table 38. Result

149

tan(0.5)

0.5463024898437905

1 row

cot()

cot() returns the cotangent of the expression.

Syntax: cot( expression )

Arguments:


Query

RETURN cot(0.5)

The cotangent of 0.5.

Table 39. Result

cot(0.5)

1.830487721712452

1 row

asin()

asin() returns the arcsine of the expression, in radians.

Syntax: asin( expression )

Arguments:


Query

RETURN asin(0.5)

The arcsine of 0.5.

Table 40. Result

asin(0.5)

0.5235987755982989

1 row

acos()

acos() returns the arccosine of the expression, in radians.

Syntax: abs( expression )

Arguments:

150


Query

RETURN acos(0.5)

The arccosine of 0.5.

Table 41. Result

acos(0.5)

1.0471975511965979

1 row

atan()

atan() returns the arctangent of the expression, in radians.

Syntax: atan( expression )

Arguments:


Query

RETURN atan(0.5)

The arctangent of 0.5.

Table 42. Result

atan(0.5)

0.4636476090008061

1 row

atan2()

atan2() returns the arctangent2 of a set of coordinates, in radians.

Syntax: atan2( expression1, expression2 )

Arguments:

• expression1: A numeric expression for y that represents the angle in radians.

• expression2: A numeric expression for x that represents the angle in radians.

Query

RETURN atan2(0.5, 0.6)

The arctangent2 of 0.5 and 0.6.

Table 43. Result

151

atan2(0.5, 0.6)

0.6947382761967033

1 row

pi()

pi() returns the mathematical constant pi.

Syntax: pi()

Arguments:

Query

RETURN pi()

The constant pi is returned.

Table 44. Result

pi()

3.141592653589793

1 row

degrees()

degrees() converts radians to degrees.

Syntax: degrees( expression )

Arguments:


Query

RETURN degrees(3.14159)

The number of degrees in something close to pi.

Table 45. Result

degrees(3.14159)

179.99984796050427

1 row

radians()

radians() returns the common logarithm (base 10) of the expression.

Syntax: radians( expression )

Arguments:

• expression: A numeric expression that represents the angle in degrees.

152

Query

RETURN radians(180)

The number of radians in 180 degrees is returned (pi).

Table 46. Result

radians(180)

3.141592653589793

1 row

haversin()

haversin() returns half the versine of the expression.

Syntax: haversin( expression )

Arguments:


Query

RETURN haversin(0.5)

The haversine of 0.5 is returned.

Table 47. Result

haversin(0.5)

0.06120871905481362

1 row

Spherical distance using the haversin function

The haversin() function may be used to compute the distance on the surface of a sphere betweentwo points (each given by their latitude and longitude). In this example the spherical distance (in km)between Berlin in Germany (at lat 52.5, lon 13.4) and San Mateo in California (at lat 37.5, lon -122.3) iscalculated using an average earth radius of 6371 km.

Query

CREATE (ber:City { lat: 52.5, lon: 13.4 }),(sm:City { lat: 37.5, lon: -122.3 })RETURN 2 * 6371 * asin(sqrt(haversin(radians(sm.lat - ber.lat))+ cos(radians(sm.lat))*cos(radians(ber.lat))* haversin(radians(sm.lon - ber.lon)))) AS dist

The estimated distance between Berlin and San Mateo is returned.

Table 48. Result

dist

9129.969740051658

153

dist

1 rowNodes created: 2Properties set: 4Labels added: 2

10.5. String functions

These functions all operate on string expressions only, and will return an error if used on any othervalues. The exception to this rule is toString(), which also accepts numbers and booleans.

See also String operators.

bar, foo



KNOWS


KNOWS


KNOWS

Spouse


MARRIEDKNOWS

Figure 21. Graph

10.5.1. replace()

replace() returns a string with the search string replaced by the replace string. It replaces alloccurrences.

Syntax: replace( original, search, replace )

Arguments:

• original: An expression that returns a string

• search: An expression that returns a string to search for

• replace: An expression that returns the string to replace the search string with

Query

RETURN replace("hello", "l", "w")

Result

+----------------------------+| replace("hello", "l", "w") |+----------------------------+| "hewwo" |+----------------------------+1 row

154

10.5.2. substring()

substring() returns a substring of the original, with a 0-based index start and length. If length isomitted, it returns a substring from start until the end of the string.

Syntax: substring( original, start [, length] )

Arguments:


• start: An expression that returns a positive number

• length: An expression that returns a positive number

Query

RETURN substring("hello", 1, 3), substring("hello", 2)

Result

+--------------------------------------------------+| substring("hello", 1, 3) | substring("hello", 2) |+--------------------------------------------------+| "ell" | "llo" |+--------------------------------------------------+1 row

10.5.3. left()

left() returns a string containing the left n characters of the original string.

Syntax: left( original, length )

Arguments:


• n: An expression that returns a positive number

Query

RETURN left("hello", 3)

Result

+------------------+| left("hello", 3) |+------------------+| "hel" |+------------------+1 row

10.5.4. right()

right() returns a string containing the right n characters of the original string.

Syntax: right( original, length )

Arguments:

155


• n: An expression that returns a positive number

Query

RETURN right("hello", 3)

Result

+-------------------+| right("hello", 3) |+-------------------+| "llo" |+-------------------+1 row

10.5.5. ltrim()

ltrim() returns the original string with whitespace removed from the left side.

Syntax: ltrim( original )

Arguments:


Query

RETURN ltrim(" hello")

Result

+-------------------+| ltrim(" hello") |+-------------------+| "hello" |+-------------------+1 row

10.5.6. rtrim()

rtrim() returns the original string with whitespace removed from the right side.

Syntax: rtrim( original )

Arguments:


Query

RETURN rtrim("hello ")

156

Result

+-------------------+| rtrim("hello ") |+-------------------+| "hello" |+-------------------+1 row

10.5.7. trim()

trim() returns the original string with whitespace removed from both sides.

Syntax: trim( original )

Arguments:


Query

RETURN trim(" hello ")

Result

+---------------------+| trim(" hello ") |+---------------------+| "hello" |+---------------------+1 row

10.5.8. lower()

lower() returns the original string in lowercase.

Syntax: lower( original )

Arguments:


Query

RETURN lower("HELLO")

Result

+----------------+| lower("HELLO") |+----------------+| "hello" |+----------------+1 row

10.5.9. upper()

upper() returns the original string in uppercase.

157

Syntax: upper( original )

Arguments:


Query

RETURN upper("hello")

Result

+----------------+| upper("hello") |+----------------+| "HELLO" |+----------------+1 row

10.5.10. split()

split() returns the sequence of strings witch are delimited by split patterns.

Syntax: split( original, splitPattern )

Arguments:


• splitPattern: The string to split the original string with

Query

RETURN split("one,two", ",")

Result

+-----------------------+| split("one,two", ",") |+-----------------------+| ["one","two"] |+-----------------------+1 row

10.5.11. reverse()

reverse() returns the original string reversed.

Syntax: reverse( original )

Arguments:


Query

RETURN reverse("anagram")

158

Result

+--------------------+| reverse("anagram") |+--------------------+| "margana" |+--------------------+1 row

10.5.12. toString()

toString() converts the argument to a string. It converts integral and floating point numbers andbooleans to strings, and if called with a string will leave it unchanged.

Syntax: toString( expression )

Arguments:

• expression: An expression that returns a number, a boolean, or a string

Query

RETURN toString(11.5), toString("already a string"), toString(TRUE )

Result

+----------------------------------------------------------------+| toString(11.5) | toString("already a string") | toString(true) |+----------------------------------------------------------------+| "11.5" | "already a string" | "true" |+----------------------------------------------------------------+1 row

159

Chapter 11. SchemaNeo4j 2.0 introduced an optional schema for the graph, based around the concept of labels. Labelsare used in the specification of indexes, and for defining constraints on the graph. Together, indexesand constraints are the schema of the graph. Cypher includes data definition language (DDL)statements for manipulating the schema.

11.1. Indexes

A database index is a redundant copy of information in the database for the purpose of makingretrieving said data more efficient. This comes at the cost of additional storage space and slowerwrites, so deciding what to index and what not to index is an important and often non-trivial task.

Cypher allows the creation of indexes over a property for all nodes that have a given label. Once anindex has been created, it will automatically be managed and kept up to date by the databasewhenever the graph is changed. Neo4j will automatically pick up and start using the index once it hasbeen created and brought online.

11.1.1. Create an index

To create an index on a property for all nodes that have a label, use CREATE INDEX ON. Note that theindex is not immediately available, but will be created in the background.

Query

CREATE INDEX ON :Person(name)

Result


11.1.2. Drop an index

To drop an index on all nodes that have a label and property combination, use the DROP INDEXclause.

Query

DROP INDEX ON :Person(name)

Result

+-------------------+| No data returned. |+-------------------+Indexes removed: 1

11.1.3. Use index

There is usually no need to specify which indexes to use in a query, Cypher will figure that out by itself.For example the query below will use the Person(name) index, if it exists. If you want Cypher to usespecific indexes, you can enforce it using hints. See Using.

160

Query

MATCH (person:Person { name: 'Andres' })RETURN person

Query Plan

+-----------------+----------------+------+---------+-----------+---------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+---------------+| +ProduceResults | 1 | 1 | 0 | person | person || | +----------------+------+---------+-----------+---------------+| +NodeIndexSeek | 1 | 1 | 2 | person | :Person(name) |+-----------------+----------------+------+---------+-----------+---------------+

Total database accesses: 2

11.1.4. Use index with WHERE using equality

Indexes are also automatically used for equality comparisons of an indexed property in the WHEREclause. If you want Cypher to use specific indexes, you can enforce it using hints. See Using.

Query

MATCH (person:Person)WHERE person.name = 'Andres'RETURN person

Query Plan



11.1.5. Use index with WHERE using inequality

Indexes are also automatically used for inequality (range) comparisons of an indexed property in theWHERE clause. If you want Cypher to use specific indexes, you can enforce it using hints. See Using.

Query

MATCH (person:Person)WHERE person.name > 'B'RETURN person

Query Plan

+-----------------------+----------------+------+---------+-----------+---------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------------+----------------+------+---------+-----------+---------------------------------+| +ProduceResults | 10 | 1 | 0 | person | person || | +----------------+------+---------+-----------+---------------------------------+| +NodeIndexSeekByRange | 10 | 1 | 2 | person | :Person(name) > { AUTOSTRING0} |+-----------------------+----------------+------+---------+-----------+---------------------------------+


161

11.1.6. Use index with IN

The IN predicate on person.name in the following query will use the Person(name) index, if it exists. Ifyou want Cypher to use specific indexes, you can enforce it using hints. See Using.

Query

MATCH (person:Person)WHERE person.name IN ['Andres', 'Mark']RETURN person

Query Plan



11.1.7. Use index with STARTS WITH

The STARTS WITH predicate on person.name in the following query will use the Person(name) index, if itexists.

Query

MATCH (person:Person)WHERE person.name STARTS WITH 'And'RETURN person

Query Plan

+-----------------------+----------------+------+---------+-----------+-------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+-----------------------+----------------+------+---------+-----------+-------------------------------------------+| +ProduceResults | 26 | 1 | 0 | person | person|| | +----------------+------+---------+-----------+-------------------------------------------+| +NodeIndexSeekByRange | 26 | 1 | 2 | person | :Person(name STARTS WITH {AUTOSTRING0}) |+-----------------------+----------------+------+---------+-----------+-------------------------------------------+


11.1.8. Use index when checking for the existence of a property

The has(p.name) predicate in the following query will use the Person(name) index, if it exists.

Query

MATCH (p:Person)WHERE exists(p.name)RETURN p

162

Query Plan

+-----------------+----------------+------+---------+-----------+---------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+---------------+| +ProduceResults | 2 | 2 | 0 | p | p || | +----------------+------+---------+-----------+---------------+| +NodeIndexScan | 2 | 2 | 3 | p | :Person(name) |+-----------------+----------------+------+---------+-----------+---------------+


11.2. Constraints

Neo4j helps enforce data integrity with the use of constraints. Constraints can be applied to

either nodes or relationships. Unique node property constraints can be created, as well as

node and relationship property existence constraints.

You can use unique property constraints to ensure that property values are unique for all nodes witha specific label. Unique constraints do not mean that all nodes have to have a unique value for theproperties — nodes without the property are not subject to this rule.

You can use property existence constraints to ensure that a property exists for all nodes with aspecific label or for all relationships with a specific type. All queries that try to create new nodes orrelationships without the property, or queries that try to remove the mandatory property will now fail.

Property existence constraints are only available in the Neo4j Enterprise Edition.Note that databases with property existence constraints cannot be opened usingNeo4j Community Edition.

You can have multiple constraints for a given label and you can also combine unique and propertyexistence constraints on the same property.

Remember that adding constraints is an atomic operation that can take a while — all existing data hasto be scanned before Neo4j can turn the constraint ``on''.

Note that adding a unique property constraint on a property will also add an index on that property,so you cannot add such an index separately. Cypher will use that index for lookups just like otherindexes. If you drop a unique property constraint and still want an index on the property, you will haveto create the index.

The existing constraints can be listed using the REST API, see [rest-api-schema-constraints].

11.2.1. Unique node property constraints

Create uniqueness constraint

To create a constraint that makes sure that your database will never contain more than one node witha specific label and one property value, use the IS UNIQUE syntax.

Query

CREATE CONSTRAINT ON (book:Book) ASSERT book.isbn IS UNIQUE

163

Result

+-------------------+| No data returned. |+-------------------+Unique constraints added: 1

Drop uniqueness constraint

By using DROP CONSTRAINT, you remove a constraint from the database.

Query

DROP CONSTRAINT ON (book:Book) ASSERT book.isbn IS UNIQUE

Result

+-------------------+| No data returned. |+-------------------+Unique constraints removed: 1

Create a node that complies with unique property constraints

Create a Book node with an isbn that isn’t already in the database.

Query

CREATE (book:Book { isbn: '1449356265', title: 'Graph Databases' })

Result


Create a node that breaks a unique property constraint

Create a Book node with an isbn that is already used in the database.

Query


In this case the node isn’t created in the graph.

Error message

Node 0 already exists with label Book and property "isbn"=[1449356265]

Failure to create a unique property constraint due to conflicting nodes

Create a unique property constraint on the property isbn on nodes with the Book label when there aretwo nodes with the same isbn.

164

Query

CREATE CONSTRAINT ON (book:Book) ASSERT book.isbn IS UNIQUE

In this case the constraint can’t be created because it is violated by existing data. We may choose touse Indexes instead or remove the offending nodes and then re-apply the constraint.

Error message

Unable to create CONSTRAINT ON ( book:Book ) ASSERT book.isbn IS UNIQUE:Multiple nodes with label `Book` have property `isbn` = '1449356265': node(0) node(1)

11.2.2. Node property existence constraints

Create node property existence constraint

To create a constraint that makes sure that all nodes with a certain label have a certain property, usethe ASSERT exists(variable.propertyName) syntax.

Query

CREATE CONSTRAINT ON (book:Book) ASSERT exists(book.isbn)

Result

+-------------------+| No data returned. |+-------------------+Property existence constraints added: 1

Drop node property existence constraint

By using DROP CONSTRAINT, you remove a constraint from the database.

Query

DROP CONSTRAINT ON (book:Book) ASSERT exists(book.isbn)

Result

+-------------------+| No data returned. |+-------------------+Property existence constraints removed: 1

Create a node that complies with property existence constraints

Create a Book node with an existing isbn property.

Query


165

Result


Create a node that breaks a property existence constraint

Trying to create a Book node without an isbn property, given a property existence constraint on:Book(isbn).

Query

CREATE (book:Book { title: 'Graph Databases' })

In this case the node isn’t created in the graph.

Error message

Node 1 with label "Book" must have the property "isbn" due to a constraint

Removing an existence constrained node property

Trying to remove the isbn property from an existing node book, given a property existence constrainton :Book(isbn).

Query

MATCH (book:Book { title: 'Graph Databases' })REMOVE book.isbn

In this case the property is not removed.

Error message

Node 0 with label "Book" must have the property "isbn" due to a constraint

Failure to create a node property existence constraint due to existing node

Create a constraint on the property isbn on nodes with the Book label when there already exists anode without an isbn.

Query

CREATE CONSTRAINT ON (book:Book) ASSERT exists(book.isbn)

In this case the constraint can’t be created because it is violated by existing data. We may choose toremove the offending nodes and then re-apply the constraint.

Error message

Unable to create CONSTRAINT ON ( book:Book ) ASSERT exists(book.isbn):Node(0) with label `Book` has no value for property `isbn`

166

11.2.3. Relationship property existence constraints

Create relationship property existence constraint

To create a constraint that makes sure that all relationships with a certain type have a certainproperty, use the ASSERT exists(variable.propertyName) syntax.

Query

CREATE CONSTRAINT ON ()-[like:LIKED]-() ASSERT exists(like.day)

Result

+-------------------+| No data returned. |+-------------------+Property existence constraints added: 1

Drop relationship property existence constraint

To remove a constraint from the database, use DROP CONSTRAINT.

Query

DROP CONSTRAINT ON ()-[like:LIKED]-() ASSERT exists(like.day)

Result

+-------------------+| No data returned. |+-------------------+Property existence constraints removed: 1

Create a relationship that complies with property existence constraints

Create a LIKED relationship with an existing day property.

Query

CREATE (user:User)-[like:LIKED { day: 'yesterday' }]->(book:Book)

Result

+-------------------+| No data returned. |+-------------------+Nodes created: 2Relationships created: 1Properties set: 1Labels added: 2

Create a relationship that breaks a property existence constraint

Trying to create a LIKED relationship without a day property, given a property existence constraint:LIKED(day).

167

Query

CREATE (user:User)-[like:LIKED]->(book:Book)

In this case the relationship isn’t created in the graph.

Error message

Relationship 1 with type "LIKED" must have the property "day" due to a constraint

Removing an existence constrained relationship property

Trying to remove the day property from an existing relationship like of type LIKED, given a propertyexistence constraint :LIKED(day).

Query

MATCH (user:User)-[like:LIKED]->(book:Book)REMOVE like.day

In this case the property is not removed.

Error message

Relationship 0 with type "LIKED" must have the property "day" due to a constraint

Failure to create a relationship property existence constraint due to existingrelationship

Create a constraint on the property day on relationships with the LIKED type when there already existsa relationship without a property named day.

Query

CREATE CONSTRAINT ON ()-[like:LIKED]-() ASSERT exists(like.day)

In this case the constraint can’t be created because it is violated by existing data. We may choose toremove the offending relationships and then re-apply the constraint.

Error message

Unable to create CONSTRAINT ON ()-[ liked:LIKED ]-() ASSERT exists(liked.day):Relationship(0) with type `LIKED` has no value for property `day`

11.3. Statistics

When you issue a Cypher query, it gets compiled to an execution plan (see Execution Plans) that canrun and answer your question. To produce an efficient plan for your query, Neo4j needs informationabout your database, such as the schema — what indexes and constraints do exist? Neo4j will also usestatistical information it keeps about your database to optimize the execution plan. With thisinformation, Neo4j can decide which access pattern leads to the best performing plans.

The statistical information that Neo4j keeps is:

1. The number of nodes with a certain label.

168

2. Selectivity per index.

3. The number of relationships by type.

4. The number of relationships by type, ending or starting from a node with a specific label.

Neo4j keeps the statistics up to date in two different ways. For label counts for example, the numberis updated whenever you set or remove a label from a node. For indexes, Neo4j needs to scan the fullindex to produce the selectivity number. Since this is potentially a very time-consuming operation,these numbers are collected in the background when enough data on the index has been changed.

11.3.1. Configuration options

Execution plans are cached and will not be replanned until the statistical information used to producethe plan has changed. The following configuration options allows you to control how sensitivereplanning should be to updates of the database.

dbms.index_sampling.background_enabled

Controls whether indexes will automatically be re-sampled when they have been updated enough.The Cypher query planner depends on accurate statistics to create efficient plans, so it is importantit is kept up to date as the database evolves.

If background sampling is turned off, make sure to trigger manual samplingwhen data has been updated.

dbms.index_sampling.update_percentage

Controls how large portion of the index has to have been updated before a new sampling run istriggered.

cypher.statistics_divergence_threshold

Controls how much the above statistical information is allowed to change before an execution planis considered stale and has to be replanned. If the relative change in any of statistics is larger thanthis threshold, the plan will be thrown away and a new one will be created. A threshold of 0.0means always replan, and a value of 1.0 means never replan.

11.3.2. Managing statistics from the shell

Usage:

schema sample -a

will sample all indexes.

schema sample -l Person -p name

will sample the index for label Person on property name (if existing).

schema sample -a -f

will force a sample of all indexes.

schema sample -f -l :Person -p name

will force sampling of a specific index.

169

Chapter 12. Query TuningNeo4j works very hard to execute queries as fast as possible.

However, when optimizing for maximum query execution performance, it may be helpful to rephrasequeries using knowledge about the domain and the application.

The overall goal of manual query performance optimization is to ensure that only necessary data isretrieved from the graph. At least data should get filtered out as early as possible in order to reducethe amount of work that has to be done at later stages of query execution. This also goes for whatgets returned: avoid returning whole nodes and relationships — instead, pick the data you need andreturn only that. You should also make sure to set an upper limit on variable length patterns, so theydon’t cover larger portions of the dataset than needed.

Each Cypher query gets optimized and transformed into an execution plan by the Cypher executionengine. To minimize the resources used for this, make sure to use parameters instead of literals whenpossible. This allows Cypher to re-use your queries instead of having to parse and build new executionplans.

To read more about the execution plan operators mentioned in this chapter, see Execution Plans.

12.1. How are queries executed?

Each query is turned into an execution plan by something called the execution planner. The executionplan tells Neo4j which operations to perform when executing the query. Two different executionplanning strategies are included in Neo4j:

Rule

This planner has rules that are used to produce execution plans. The planner considers availableindexes, but does not use statistical information to guide the query compilation.

Cost

This planner uses the statistics service in Neo4j to assign cost to alternative plans and picks thecheapest one. While this should lead to superior execution plans in most cases, it is still underdevelopment.

By default, Neo4j 3.0 will use the cost planner for some queries, but not all. You can force it to use aspecific planner by using the cypher.planner configuration setting (see [config_cypher.planner]), or byprepending your query with CYPHER planner=cost or CYPHER planner=rule. Neo4j might still not use theplanner you selected — not all queries are solvable by the cost planner at this point.

You can see which planner was used by looking at the execution plan.

When Cypher is building execution plans, it looks at the schema to see if it can findindexes it can use. These index decisions are only valid until the schema changes,so adding or removing indexes leads to the execution plan cache being flushed.

12.2. How do I profile a query?

There are two options to choose from when you want to analyze a query by looking at its executionplan:

EXPLAIN

If you want to see the execution plan but not run the statement, prepend your Cypher statementwith EXPLAIN. The statement will always return an empty result and make no changes to thedatabase.

170

PROFILE

If you want to run the statement and see which operators are doing most of the work, use PROFILE.This will run your statement and keep track of how many rows pass through each operator, andhow much each operator needs to interact with the storage layer to retrieve the necessary data.Please note that profiling your query uses more resources, so you should not profile unless you areactively working on a query.

See Execution Plans for a detailed explanation of each of the operators contained in an executionplan.

Being explicit about what types and labels you expect relationships and nodes tohave in your query helps Neo4j use the best possible statistical information, whichleads to better execution plans. This means that when you know that a relationshipcan only be of a certain type, you should add that to the query. The same goes forlabels, where declaring labels on both the start and end nodes of a relationshiphelps Neo4j find the best way to execute the statement.

12.3. Basic query tuning example

We’ll start with a basic example to help you get the hang of profiling queries. The following exampleswill use a movies data set.

Let’s start by importing the data:

LOAD CSV WITH HEADERS FROM "http://neo4j.com/docs/3.0.1/csv/query-tuning/movies.csv" AS lineMERGE (m:Movie { title:line.title })ON CREATE SET m.released = toInt(line.released), m.tagline = line.tagline

LOAD CSV WITH HEADERS FROM 'http://neo4j.com/docs/3.0.1/csv/query-tuning/actors.csv' AS lineMATCH (m:Movie { title:line.title })MERGE (p:Person { name:line.name })ON CREATE SET p.born = toInt(line.born)MERGE (p)-[:ACTED_IN { roles:split(line.roles,";")}]->(m)

LOAD CSV WITH HEADERS FROM 'http://neo4j.com/docs/3.0.1/csv/query-tuning/directors.csv' AS lineMATCH (m:Movie { title:line.title })MERGE (p:Person { name:line.name })ON CREATE SET p.born = toInt(line.born)MERGE (p)-[:DIRECTED]->(m)

Let’s say we want to write a query to find Tom Hanks. The naive way of doing this would be to writethe following:

MATCH (p { name:"Tom Hanks" })RETURN p

This query will find the Tom Hanks node but as the number of nodes in the database increase it willbecome slower and slower. We can profile the query to find out why that is.

You can learn more about the options for profiling queries in How do I profile a query? but in this casewe’re going to prefix our query with PROFILE:

PROFILEMATCH (p { name:"Tom Hanks" })RETURN p

171

Compiler CYPHER 3.0

Planner COST

Runtime INTERPRETED

+-----------------+----------------+------+---------+-----------+---------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+---------------------------+| +ProduceResults | 16 | 1 | 0 | p | p || | +----------------+------+---------+-----------+---------------------------+| +Filter | 16 | 1 | 163 | p | p.name == { AUTOSTRING0} || | +----------------+------+---------+-----------+---------------------------+| +AllNodesScan | 163 | 163 | 164 | p | |+-----------------+----------------+------+---------+-----------+---------------------------+


The first thing to keep in mind when reading execution plans is that you need to read from the bottomup.

In that vein, starting from the last row, the first thing we notice is that the value in the Rows columnseems high given there is only one node with the name property Tom Hanks in the database. If we lookacross to the Operator column we’ll see that AllNodesScan has been used which means that the queryplanner scanned through all the nodes in the database.

Moving up to the previous row we see the Filter operator which will check the name property on each ofthe nodes passed through by AllNodesScan.

This seems like an inefficient way of finding Tom Hanks given that we are looking at many nodes thataren’t even people and therefore aren’t what we’re looking for.

The solution to this problem is that whenever we’re looking for a node we should specify a label tohelp the query planner narrow down the search space. For this query we’d need to add a Person label.

MATCH (p:Person { name:"Tom Hanks" })RETURN p

This query will be faster than the first one but as the number of people in our database increase weagain notice that the query slows down.

Again we can profile the query to work out why:

PROFILEMATCH (p:Person { name:"Tom Hanks" })RETURN p

Compiler CYPHER 3.0

Planner COST

Runtime INTERPRETED

+------------------+----------------+------+---------+-----------+---------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+------------------+----------------+------+---------+-----------+---------------------------+| +ProduceResults | 13 | 1 | 0 | p | p || | +----------------+------+---------+-----------+---------------------------+| +Filter | 13 | 1 | 125 | p | p.name == { AUTOSTRING0} || | +----------------+------+---------+-----------+---------------------------+| +NodeByLabelScan | 125 | 125 | 126 | p | :Person |+------------------+----------------+------+---------+-----------+---------------------------+


172

This time the Rows value on the last row has reduced so we’re not scanning some nodes that we werebefore which is a good start. The NodeByLabelScan operator indicates that we achieved this by firstdoing a linear scan of all the Person nodes in the database.

Once we’ve done that we again scan through all those nodes using the Filter operator, comparingthe name property of each one.

This might be acceptable in some cases but if we’re going to be looking up people by name frequentlythen we’ll see better performance if we create an index on the name property for the Person label:

CREATE INDEX ON :Person(name)

Now if we run the query again it will run more quickly:

MATCH (p:Person { name:"Tom Hanks" })RETURN p

Let’s profile the query to see why that is:

PROFILEMATCH (p:Person { name:"Tom Hanks" })RETURN p

Compiler CYPHER 3.0

Planner COST

Runtime INTERPRETED

+-----------------+----------------+------+---------+-----------+---------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+---------------+| +ProduceResults | 1 | 1 | 0 | p | p || | +----------------+------+---------+-----------+---------------+| +NodeIndexSeek | 1 | 1 | 2 | p | :Person(name) |+-----------------+----------------+------+---------+-----------+---------------+


Our execution plan is down to a single row and uses the Node Index Seek operator which does aschema index seek (see Indexes) to find the appropriate node.

12.4. Using

The USING clause is used to influence the decisions of the planner when building an execution

plan for a query.

Forcing planner behavior is an advanced feature, and should be used with cautionby experienced developers and/or database administrators only, as it may causequeries to perform poorly.

12.4.1. Introduction

When executing a query, Neo4j needs to decide where in the query graph to start matching. This isdone by looking at the MATCH clause and the WHERE conditions and using that information to find usefulindexes, or other starting points.

173

However, the selected index might not always be the best choice. Sometimes multiple indexes arepossible candidates, and the query planner picks the wrong one from a performance point of view.And in some circumstances (albeit rarely) it is better not to use an index at all.

You can force Neo4j to use a specific starting point through the USING clause. This is called giving aplanner hint. There are three types of planner hints: index hints, scan hints, and join hints.

You cannot use planner hints if your query has a START clause.


Scientist

name = 'Liskov'born = 1939

Science

name = 'Computer Science'

RESEARCHED

Scientist

name = 'Wing'born = 1956

KNOWS

RESEARCHED

Science

name = 'Engineering'

RESEARCHED

Scientist

name = 'Conway'born = 1938

RESEARCHED

Science

name = 'Chemistry'

Scientist

name = 'Curie'born = 1867

RESEARCHED

Scientist

name = 'Arden'

RESEARCHED

Scientist

name = 'Franklin'

RESEARCHED

Scientist

name = 'Harrison'

RESEARCHED

Figure 22. Graph

Query

MATCH (liskov:Scientist { name:'Liskov' })-[:KNOWS]->(wing:Scientist)-[:RESEARCHED]->(cs:Science {name:'Computer Science' })<-[:RESEARCHED]-(conway:Scientist { name: 'Conway' })RETURN 1 AS column

The following query will be used in some of the examples on this page. It has intentionally beenconstructed in such a way that the statistical information will be inaccurate for the particular subgraphthat this query matches. For this reason, it can be improved by supplying planner hints.

174

Query plan

+------------------+----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables| Other |+------------------+----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +ProduceResults | 0 | 1 | 0 | column| column || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Projection | 0 | 1 | 0 | column -- anon[122], anon[41], anon[68], conway,cs, liskov, wing | { AUTOINT3} || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Filter | 0 | 1 | 2 | anon[122], anon[41], anon[68], conway, cs, liskov,wing | liskov.name == { AUTOSTRING0} AND liskov:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Expand(All) | 0 | 1 | 3 | anon[41], liskov -- anon[122], anon[68], conway,cs, wing | (wing)<-[:KNOWS]-(liskov) || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Filter | 1 | 2 | 2 | anon[122], anon[68], conway, cs, wing| NOT(anon[122] == anon[68]) AND wing:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Expand(All) | 1 | 3 | 4 | anon[68], wing -- anon[122], conway, cs| (cs)<-[:RESEARCHED]-(wing) || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Filter | 0 | 1 | 4 | anon[122], conway, cs| conway.name == { AUTOSTRING2} AND conway:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Expand(All) | 3 | 3 | 4 | anon[122], conway -- cs| (cs)<-[:RESEARCHED]-(conway) || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +Filter | 1 | 1 | 3 | cs| cs.name == { AUTOSTRING1} || | +----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+| +NodeByLabelScan | 3 | 3 | 4 | cs| :Science |+------------------+----------------+------+---------+-------------------------------------------------------------------+-----------------------------------------------------+


12.4.2. Index hints

Index hints are used to specify which index, if any, the planner should use as a starting point. This canbe beneficial in cases where the index statistics are not accurate for the specific values that the queryat hand is known to use, which would result in the planner picking a non-optimal index. To supply anindex hint, use USING INDEX variable:Label(property) after the applicable MATCH clause.

It is possible to supply several index hints, but keep in mind that several starting points will require theuse of a potentially expensive join later in the query plan.

175

Query using an index hint

The query above will not naturally pick an index to solve the plan. This is because the graph is verysmall, and label scans are faster for small databases. In general, however, query performance isranked by the dbhit metric, and we see that using an index is slightly better for this query.

Query

MATCH (liskov:Scientist { name:'Liskov' })-[:KNOWS]->(wing:Scientist)-[:RESEARCHED]->(cs:Science {name:'Computer Science' })<-[:RESEARCHED]-(conway:Scientist { name: 'Conway' })USING INDEX liskov:Scientist(name)RETURN liskov.born AS column

Returns the year Barbara Liskov was born.

Query plan

+-----------------+----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables| Other |+-----------------+----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +ProduceResults | 0 | 1 | 0 | column| column || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Projection | 0 | 1 | 1 | column -- anon[122], anon[41], anon[68], conway, cs,liskov, wing | liskov.born || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Filter | 0 | 1 | 3 | anon[122], anon[41], anon[68], conway, cs, liskov,wing | NOT(anon[122] == anon[68]) AND conway.name == { AUTOSTRING2} AND conway:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Expand(All) | 0 | 3 | 4 | anon[122], conway -- anon[41], anon[68], cs, liskov,wing | (cs)<-[:RESEARCHED]-(conway) || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Filter | 0 | 1 | 4 | anon[41], anon[68], cs, liskov, wing| cs:Science AND cs.name == { AUTOSTRING1} || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Expand(All) | 0 | 2 | 3 | anon[68], cs -- anon[41], liskov, wing| (wing)-[:RESEARCHED]->(cs) || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Filter | 0 | 1 | 1 | anon[41], liskov, wing| wing:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +Expand(All) | 0 | 1 | 2 | anon[41], wing -- liskov| (liskov)-[:KNOWS]->(wing) || | +----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+| +NodeIndexSeek | 1 | 1 | 2 | liskov| :Scientist(name) |+-----------------+----------------+------+---------+-------------------------------------------------------------------+------------------------------------------------------------------------------------+


176

Query using multiple index hints

Supplying one index hint changed the starting point of the query, but the plan is still linear, meaning itonly has one starting point. If we give the planner yet another index hint, we force it to use twostarting points, one at each end of the match. It will then join these two branches using a joinoperator.

Query

MATCH (liskov:Scientist { name:'Liskov' })-[:KNOWS]->(wing:Scientist)-[:RESEARCHED]->(cs:Science {name:'Computer Science' })<-[:RESEARCHED]-(conway:Scientist { name: 'Conway' })USING INDEX liskov:Scientist(name)USING INDEX conway:Scientist(name)RETURN liskov.born AS column

Returns the year Barbara Liskov was born, using a slightly better plan.

177

Query plan

+------------------+----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables| Other |+------------------+----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +ProduceResults | 0 | 1 | 0 | column| column || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Projection | 0 | 1 | 1 | column -- anon[122], anon[41], anon[68], conway,cs, liskov, wing | liskov.born || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Filter | 0 | 1 | 0 | anon[122], anon[41], anon[68], conway, cs, liskov,wing | NOT(anon[122] == anon[68]) || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +NodeHashJoin | 0 | 1 | 0 | anon[41], anon[68], liskov, wing -- anon[122],conway, cs | cs || |\ +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +Filter | 1 | 1 | 1 | anon[122], conway, cs| cs.name == { AUTOSTRING1} || | | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +Expand(All) | 1 | 1 | 2 | anon[122], cs -- conway| (conway)-[:RESEARCHED]->(cs) || | | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +NodeIndexSeek | 1 | 1 | 2 | conway| :Scientist(name) || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Filter | 0 | 1 | 4 | anon[41], anon[68], cs, liskov, wing| cs:Science AND cs.name == { AUTOSTRING1} || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Expand(All) | 0 | 2 | 3 | anon[68], cs -- anon[41], liskov, wing| (wing)-[:RESEARCHED]->(cs) || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Filter | 0 | 1 | 1 | anon[41], liskov, wing| wing:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Expand(All) | 0 | 1 | 2 | anon[41], wing -- liskov| (liskov)-[:KNOWS]->(wing) || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +NodeIndexSeek | 1 | 1 | 2 | liskov| :Scientist(name) |+------------------+----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+


178

12.4.3. Scan hints

If your query matches large parts of an index, it might be faster to scan the label and filter out nodesthat do not match. To do this, you can use USING SCAN variable:Label after the applicable MATCHclause. This will force Cypher to not use an index that could have been used, and instead do a labelscan.

Hinting a label scan

If the best performance is to be had by scanning all nodes in a label and then filtering on that set, useUSING SCAN.

Query

MATCH (s:Scientist)USING SCAN s:ScientistWHERE s.born < 1939RETURN s.born AS column

Returns all scientists born before 1939.

Query plan

+------------------+----------------+------+---------+-------------+-------------------------------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+------------------+----------------+------+---------+-------------+-------------------------------------------------------------------+| +ProduceResults | 0 | 2 | 0 | column | column|| | +----------------+------+---------+-------------+-------------------------------------------------------------------+| +Projection | 0 | 2 | 2 | column -- s | s.born|| | +----------------+------+---------+-------------+-------------------------------------------------------------------+| +Filter | 0 | 2 | 7 | s |AndedPropertyComparablePredicates(s,s.born,s.born < { AUTOINT0}) || | +----------------+------+---------+-------------+-------------------------------------------------------------------+| +NodeByLabelScan | 7 | 7 | 8 | s | :Scientist|+------------------+----------------+------+---------+-------------+-------------------------------------------------------------------+


12.4.4. Join hints

Join hints are the most advanced type of hints, and are not used to find starting points for the queryexecution plan, but to enforce that joins are made at specified points. This implies that there has to bemore than one starting point (leaf) in the plan, in order for the query to be able to join the twobranches ascending from these leaves. Due to this nature, joins, and subsequently join hints, will forcethe planner to look for additional starting points, and in the case where there are no more good ones,potentially pick a very bad starting point. This will negatively affect query performance. In other cases,the hint might force the planner to pick a seemingly bad starting point, which in reality proves to be avery good one.

Hinting a join on a single node

In the example above using multiple index hints, we saw that the planner chose to do a join on the csnode. This means that the relationship between wing and cs was traversed in the outgoing direction,which is better statistically because the pattern ()-[:RESEARCHED]→(:Science) is more common than

179

the pattern (:Scientist)-[:RESEARCHED]→(). However, in the actual graph, the cs node only has twosuch relationships, so expanding from it will be beneficial to expanding from the wing node. We canforce the join to happen on wing instead with a join hint.

Query

MATCH (liskov:Scientist { name:'Liskov' })-[:KNOWS]->(wing:Scientist)-[:RESEARCHED]->(cs:Science {name:'Computer Science' })<-[:RESEARCHED]-(conway:Scientist { name: 'Conway' })USING INDEX liskov:Scientist(name)USING INDEX conway:Scientist(name)USING JOIN ON wingRETURN wing.born AS column

Returns the birth date of Jeanette Wing, using a slightly better plan.

180

Query plan

+------------------+----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables| Other |+------------------+----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +ProduceResults | 0 | 1 | 0 | column| column || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Projection | 0 | 1 | 1 | column -- anon[122], anon[41], anon[68], conway,cs, liskov, wing | wing.born || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +NodeHashJoin | 0 | 1 | 0 | anon[41], liskov -- anon[122], anon[68], conway,cs, wing | wing || |\ +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +Filter | 1 | 2 | 0 | anon[122], anon[68], conway, cs, wing| NOT(anon[122] == anon[68]) || | | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +Expand(All) | 1 | 3 | 4 | anon[68], wing -- anon[122], conway, cs| (cs)<-[:RESEARCHED]-(wing) || | | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +Filter | 0 | 1 | 2 | anon[122], conway, cs| cs:Science AND cs.name == { AUTOSTRING1} || | | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +Expand(All) | 1 | 1 | 2 | anon[122], cs -- conway| (conway)-[:RESEARCHED]->(cs) || | | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| | +NodeIndexSeek | 1 | 1 | 2 | conway| :Scientist(name) || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Filter | 0 | 1 | 1 | anon[41], liskov, wing| wing:Scientist || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +Expand(All) | 0 | 1 | 2 | anon[41], wing -- liskov| (liskov)-[:KNOWS]->(wing) || | +----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+| +NodeIndexSeek | 1 | 1 | 2 | liskov| :Scientist(name) |+------------------+----------------+------+---------+-------------------------------------------------------------------+-------------------------------------------+


Hinting a join on multiple nodes

The query planner can be made to produce a join between several specific points. This requires thequery to expand from the same node from several directions.

181

Query

MATCH (liskov:Scientist { name:'Liskov' })-[:KNOWS]->(wing:Scientist { name:'Wing' })-[:RESEARCHED]->(cs:Science { name:'Computer Science' })<-[:RESEARCHED]-(liskov)USING INDEX liskov:Scientist(name)USING JOIN ON liskov, csRETURN wing.born AS column

Returns the birth date of Jeanette Wing.

182

Query plan

+------------------+----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables| Other |+------------------+----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +ProduceResults | 0 | 1 | 0 | column| column || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +Projection | 0 | 1 | 1 | column -- anon[136], anon[41], anon[82], cs,liskov, wing | wing.born || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +Filter | 0 | 1 | 0 | anon[136], anon[41], anon[82], cs, liskov, wing| NOT(anon[136] == anon[82]) || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +NodeHashJoin | 0 | 1 | 0 | anon[41], anon[82], wing -- anon[136], cs, liskov| liskov, cs || |\ +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| | +Filter | 1 | 1 | 2 | anon[136], cs, liskov| cs:Science AND cs.name == { AUTOSTRING2} || | | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| | +Expand(All) | 1 | 1 | 2 | anon[136], cs -- liskov| (liskov)-[:RESEARCHED]->(cs) || | | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| | +NodeIndexSeek | 1 | 1 | 2 | liskov| :Scientist(name) || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +Filter | 0 | 1 | 4 | anon[41], anon[82], cs, liskov, wing| cs:Science AND cs.name == { AUTOSTRING2} || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +Expand(All) | 0 | 2 | 3 | anon[82], cs -- anon[41], liskov, wing| (wing)-[:RESEARCHED]->(cs) || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +Filter | 0 | 1 | 2 | anon[41], liskov, wing| wing:Scientist AND wing.name == { AUTOSTRING1} || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +Expand(All) | 0 | 1 | 2 | anon[41], wing -- liskov| (liskov)-[:KNOWS]->(wing) || | +----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+| +NodeIndexSeek | 1 | 1 | 2 | liskov| :Scientist(name) |+------------------+----------------+------+---------+-----------------------------------------------------------+-------------------------------------------------+


183

Chapter 13. Execution PlansNeo4j breaks down the work of executing a query into small pieces called operators. Each operator isresponsible for a small part of the overall query. The operators are connected together in a patterncalled a execution plan.

Each operator is annotated with statistics.

Rows

The number of rows that the operator produced. Only available if the query was profiled.

EstimatedRows

If Neo4j used the cost-based compiler you will see the estimated number of rows that will beproduced by the operator. The compiler uses this estimate to choose a suitable execution plan.

DbHits

Each operator will ask the Neo4j storage engine to do work such as retrieving or updating data. Adatabase hit is an abstract unit of this storage engine work.

See How do I profile a query? for how to view the execution plan for your query.

For a deeper understanding of how each operator works, see the relevant section. Operators aregrouped into high-level categories. Please remember that the statistics of the actual database wherethe queries run on will decide the plan used. There is no guarantee that a specific query will always besolved with the same plan.

13.1. Starting point operators

These operators find parts of the graph from which to start.

13.1.1. All Nodes Scan

Reads all nodes from the node store. The variable that will contain the nodes is seen in thearguments. If your query is using this operator, you are very likely to see performance problems onany non-trivial database.

Query

MATCH (n)RETURN n

Query Plan

+-----------------+----------------+------+---------+-----------+-------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+-------+| +ProduceResults | 35 | 35 | 0 | n | n || | +----------------+------+---------+-----------+-------+| +AllNodesScan | 35 | 35 | 36 | n | |+-----------------+----------------+------+---------+-----------+-------+


13.1.2. Directed Relationship By Id Seek

Reads one or more relationships by id from the relationship store. Produces both the relationship andthe nodes on either side.

184

Query

MATCH (n1)-[r]->()WHERE id(r)= 0RETURN r, n1

Query Plan

+-----------------------------------+----------------+------+---------+-----------------+--------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+-----------------------------------+----------------+------+---------+-----------------+--------------------------------------------+| +ProduceResults | 1 | 1 | 0 | n1, r | r, n1|| | +----------------+------+---------+-----------------+--------------------------------------------+| +DirectedRelationshipByIdSeekPipe | 1 | 1 | 1 | anon[17], n1, r |EntityByIdRhs(SingleSeekArg({ AUTOINT0})) |+-----------------------------------+----------------+------+---------+-----------------+--------------------------------------------+


13.1.3. Node by Id seek

Reads one or more nodes by id from the node store.

Query

MATCH (n)WHERE id(n)= 0RETURN n

Query Plan

+-----------------+----------------+------+---------+-----------+-------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+-------+| +ProduceResults | 1 | 1 | 0 | n | n || | +----------------+------+---------+-----------+-------+| +NodeByIdSeek | 1 | 1 | 1 | n | |+-----------------+----------------+------+---------+-----------+-------+


13.1.4. Node by label scan

Using the label index, fetches all nodes with a specific label on them from the node label index.

Query

MATCH (person:Person)RETURN person

185

Query Plan

+------------------+----------------+------+---------+-----------+---------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+------------------+----------------+------+---------+-----------+---------+| +ProduceResults | 14 | 14 | 0 | person | person || | +----------------+------+---------+-----------+---------+| +NodeByLabelScan | 14 | 14 | 15 | person | :Person |+------------------+----------------+------+---------+-----------+---------+


13.1.5. Node index seek

Finds nodes using an index seek. The node variable and the index used is shown in the arguments ofthe operator. If the index is a unique index, the operator is called NodeUniqueIndexSeek instead.

Query

MATCH (location:Location { name: "Malmo" })RETURN location

Query Plan

+-----------------+----------------+------+---------+-----------+-----------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+-----------------+| +ProduceResults | 1 | 1 | 0 | location | location || | +----------------+------+---------+-----------+-----------------+| +NodeIndexSeek | 1 | 1 | 2 | location | :Location(name) |+-----------------+----------------+------+---------+-----------+-----------------+


13.1.6. Node index range seek

Finds nodes using an index seek where the value of the property matches a given prefix string. Thisoperator can be used for STARTS WITH and comparators such as <, >, ⇐ and >=

Query

MATCH (l:Location)WHERE l.name STARTS WITH 'Lon'RETURN l

Query Plan

+-----------------------+----------------+------+---------+-----------+---------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+-----------------------+----------------+------+---------+-----------+---------------------------------------------+| +ProduceResults | 26 | 1 | 0 | l | l|| | +----------------+------+---------+-----------+---------------------------------------------+| +NodeIndexSeekByRange | 26 | 1 | 2 | l | :Location(name STARTS WITH {AUTOSTRING0}) |+-----------------------+----------------+------+---------+-----------+---------------------------------------------+


186

13.1.7. Node index contains scan

An index contains scan goes through all values stored in an index, and searches for entries containinga specific string. This is slower than an index seek, since all entries need to be examined, but stillfaster than the indirection needed by a label scan and then a property store filter.

Query

MATCH (l:Location)WHERE l.name CONTAINS 'al'RETURN l

Query Plan

+------------------------+----------------+------+---------+-----------+----------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+------------------------+----------------+------+---------+-----------+----------------------------------+| +ProduceResults | 10 | 2 | 0 | l | l|| | +----------------+------+---------+-----------+----------------------------------+| +NodeIndexContainsScan | 10 | 2 | 3 | l | :Location(name); { AUTOSTRING0}|+------------------------+----------------+------+---------+-----------+----------------------------------+


13.1.8. Node index scan

An index scan goes through all values stored in an index, and can be used to find all nodes with aparticular label having a specified property (e.g. exists(n.prop)).

Query

MATCH (l:Location)WHERE exists(l.name)RETURN l

Query Plan

+-----------------+----------------+------+---------+-----------+-----------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+-----------------+| +ProduceResults | 10 | 10 | 0 | l | l || | +----------------+------+---------+-----------+-----------------+| +NodeIndexScan | 10 | 10 | 11 | l | :Location(name) |+-----------------+----------------+------+---------+-----------+-----------------+


13.1.9. Undirected Relationship By Id Seek

Reads one or more relationships by id from the relationship store. For each relationship, two rows areproduced with start and end nodes arranged differently.

187

Query

MATCH (n1)-[r]-()WHERE id(r)= 1RETURN r, n1

Query Plan

+---------------------------------+----------------+------+---------+-----------------+-------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+---------------------------------+----------------+------+---------+-----------------+-------+| +ProduceResults | 1 | 2 | 0 | n1, r | r, n1 || | +----------------+------+---------+-----------------+-------+| +UndirectedRelationshipByIdSeek | 1 | 2 | 1 | anon[16], n1, r | |+---------------------------------+----------------+------+---------+-----------------+-------+


13.2. Expand operators

Thes operators explore the graph by expanding graph patterns.

13.2.1. Expand All

Given a start node, expand-all will follow relationships coming in or out, depending on the patternrelationship. Can also handle variable length pattern relationships.

Query

MATCH (p:Person { name: "me" })-[:FRIENDS_WITH]->(fof)RETURN fof

Query Plan

+-----------------+----------------+------+---------+--------------------+----------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+--------------------+----------------------------+| +ProduceResults | 0 | 1 | 0 | fof | fof || | +----------------+------+---------+--------------------+----------------------------+| +Expand(All) | 0 | 1 | 2 | anon[30], fof -- p | (p)-[:FRIENDS_WITH]->(fof) || | +----------------+------+---------+--------------------+----------------------------+| +NodeIndexSeek | 1 | 1 | 2 | p | :Person(name) |+-----------------+----------------+------+---------+--------------------+----------------------------+


13.2.2. Expand Into

When both the start and end node have already been found, expand-into is used to find all connectingrelationships between the two nodes.

Query

MATCH (p:Person { name: "me" })-[:FRIENDS_WITH]->(fof)-->(p)RETURN fof

188

Query Plan

+-----------------+----------------+------+---------+------------------------------+----------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+-----------------+----------------+------+---------+------------------------------+----------------------------+| +ProduceResults | 0 | 0 | 0 | fof | fof|| | +----------------+------+---------+------------------------------+----------------------------+| +Filter | 0 | 0 | 0 | anon[30], anon[53], fof, p | NOT(anon[30] ==anon[53]) || | +----------------+------+---------+------------------------------+----------------------------+| +Expand(Into) | 0 | 0 | 0 | anon[30] -- anon[53], fof, p | (p)-[:FRIENDS_WITH]->(fof) || | +----------------+------+---------+------------------------------+----------------------------+| +Expand(All) | 0 | 0 | 1 | anon[53], fof -- p | (p)<--(fof)|| | +----------------+------+---------+------------------------------+----------------------------+| +NodeIndexSeek | 1 | 1 | 2 | p | :Person(name)|+-----------------+----------------+------+---------+------------------------------+----------------------------+


13.2.3. Optional Expand All

Optional expand traverses relationships from a given node, and makes sure that predicates areevaluated before producing rows.

If no matching relationships are found, a single row with NULL for the relationship and end nodevariable is produced.

Query

MATCH (p:Person)OPTIONAL MATCH (p)-[works_in:WORKS_IN]->(l)WHERE works_in.duration > 180RETURN p, l

Query Plan

+----------------------+----------------+------+---------+------------------+------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+----------------------+----------------+------+---------+------------------+------------------------------+| +ProduceResults | 14 | 15 | 0 | l, p | p, l|| | +----------------+------+---------+------------------+------------------------------+| +OptionalExpand(All) | 14 | 15 | 44 | l, works_in -- p | (p)-[works_in:WORKS_IN]->(l)|| | +----------------+------+---------+------------------+------------------------------+| +NodeByLabelScan | 14 | 14 | 15 | p | :Person|+----------------------+----------------+------+---------+------------------+------------------------------+


189

13.3. Combining operators

The combining operators are used to piece together other operators.


Person

name = 'me'

Person

name = 'Andres'

FRIENDS_WITH

Location

name = 'London'

WORKS_INduration = 190


Person

name = 'Andreas'

FRIENDS_WITH

Country

name = 'England'

IN


Person

name = 'Mattias'


Person

name = 'Lovis'

Location

name = 'San Francisco'


Person

name = 'Pontus'

Location

name = 'Malmo'


Person

name = 'Max'

Location

name = 'New York'


Person

name = 'Konstantin'


Person

name = 'Stefan'


Location

name = 'Berlin'


Person

name = 'Mats'


Person

name = 'Petra'


Person

name = 'Craig'


Person

name = 'Steven'


Person

name = 'Chris'

Location

name = 'Madrid'


Location

name = 'Kuala Lumpur'

Location

name = 'Stockholm'

Location

name = 'Paris'

Location

name = 'Rome'

Team

name = 'Field'

Team

name = 'Engineering'id = 42

Team

name = 'Sales'

Team

name = 'Team Monads'

Team

name = 'Team Enlightened Birdmen'

Team

name = 'Team Quality'

Team

name = 'Team Rassilon'

Team

name = 'Team Executive'

Team

name = 'Team Remoting'

Team

name = 'Other'

Figure 23. Graph

13.3.1. Apply

Apply works by performing a nested loop. Every row being produced on the left-hand side of the Applyoperator will be fed to the leaf operator on the right-hand side, and then Apply will yield the combinedresults. Apply, being a nested loop, can be seen as a warning that a better plan was not found.

Query

MATCH (p:Person)-[:FRIENDS_WITH]->(f)WITH p, count(f) AS fsWHERE fs > 2OPTIONAL MATCH (p)-[:WORKS_IN]->(city)RETURN city.name

Finds all the people with more than two friends and returns the city they work in.

190

Query plan

+----------------------+----------------+----------------------------------------------+--------------------------+| Operator | Estimated Rows | Variables | Other|+----------------------+----------------+----------------------------------------------+--------------------------+| +ProduceResults | 1 | city.name | city.name|| | +----------------+----------------------------------------------+--------------------------+| +Projection | 1 | city.name -- anon[70], anon[93], city, fs, p | city.name|| | +----------------+----------------------------------------------+--------------------------+| +OptionalExpand(All) | 1 | anon[93], city -- anon[70], fs, p | (p)-[:WORKS_IN]->(city) || | +----------------+----------------------------------------------+--------------------------+| +Filter | 1 | anon[70], fs, p | anon[70]|| | +----------------+----------------------------------------------+--------------------------+| +Projection | 1 | anon[70] -- fs, p | p; fs; fs > {AUTOINT0} || | +----------------+----------------------------------------------+--------------------------+| +EagerAggregation | 1 | fs -- p | p|| | +----------------+----------------------------------------------+--------------------------+| +Expand(All) | 2 | anon[17], f -- p | (p)-[:FRIENDS_WITH]->(f) || | +----------------+----------------------------------------------+--------------------------+| +NodeByLabelScan | 14 | p | :Person|+----------------------+----------------+----------------------------------------------+--------------------------+

Total database accesses: ?

13.3.2. SemiApply

Tests for the existence of a pattern predicate. SemiApply takes a row from its child operator and feedsit to the leaf operator on the right-hand side. If the right-hand side operator tree yields at least onerow, the row from the left-hand side is yielded by the SemiApply operator. This makes SemiApply afiltering operator, used mostly for pattern predicates in queries.

Query

MATCH (p:Person)WHERE (p)-[:FRIENDS_WITH]->()RETURN p.name

Finds all the people who have friends.

191

Query plan

+------------------+----------------+-------------------------+-------------------------+| Operator | Estimated Rows | Variables | Other |+------------------+----------------+-------------------------+-------------------------+| +ProduceResults | 11 | p.name | p.name || | +----------------+-------------------------+-------------------------+| +Projection | 11 | p.name -- p | p.name || | +----------------+-------------------------+-------------------------+| +SemiApply | 11 | p | || |\ +----------------+-------------------------+-------------------------+| | +Expand(All) | 2 | anon[27], anon[45] -- p | (p)-[:FRIENDS_WITH]->() || | | +----------------+-------------------------+-------------------------+| | +Argument | 14 | p | || | +----------------+-------------------------+-------------------------+| +NodeByLabelScan | 14 | p | :Person |+------------------+----------------+-------------------------+-------------------------+


13.3.3. AntiSemiApply

Tests for the existence of a pattern predicate. SemiApply takes a row from its child operator and feedsit to the leaf operator on the right-hand side. If the right-hand side operator tree yields at least onerow, the row from the left-hand side is yielded by the SemiApply operator. This makes SemiApply afiltering operator, used mostly for pattern predicates in queries.

Query

MATCH (me:Person { name: "me" }),(other:Person)WHERE NOT (me)-[:FRIENDS_WITH]->(other)RETURN other.name

Finds the names of all the people who are not my friends.

Query plan

+--------------------+----------------+-------------------------+-------------------------------+| Operator | Estimated Rows | Variables | Other |+--------------------+----------------+-------------------------+-------------------------------+| +ProduceResults | 4 | other.name | other.name || | +----------------+-------------------------+-------------------------------+| +Projection | 4 | other.name -- me, other | other.name || | +----------------+-------------------------+-------------------------------+| +AntiSemiApply | 4 | me, other | || |\ +----------------+-------------------------+-------------------------------+| | +Expand(Into) | 0 | anon[62] -- me, other | (me)-[:FRIENDS_WITH]->(other) || | | +----------------+-------------------------+-------------------------------+| | +Argument | 14 | me, other | || | +----------------+-------------------------+-------------------------------+| +CartesianProduct | 14 | me -- other | || |\ +----------------+-------------------------+-------------------------------+| | +NodeByLabelScan | 14 | other | :Person || | +----------------+-------------------------+-------------------------------+| +NodeIndexSeek | 1 | me | :Person(name) |+--------------------+----------------+-------------------------+-------------------------------+


13.3.4. LetSemiApply

Tests for the existence of a pattern predicate. When a query contains multiple pattern predicatesLetSemiApply will be used to evaluate the first of these. It will record the result of evaluating thepredicate but will leave any filtering to a another operator.

192

Query

MATCH (other:Person)WHERE (other)-[:FRIENDS_WITH]->() OR (other)-[:WORKS_IN]->()RETURN other.name

Finds the names of all the people who have a friend or who work somewhere. The LetSemiApplyoperator will be used to check for the existence of the FRIENDS_WITH relationship from each person.

Query plan

+--------------------+----------------+-------------------------------+-----------------------------+| Operator | Estimated Rows | Variables | Other |+--------------------+----------------+-------------------------------+-----------------------------+| +ProduceResults | 13 | other.name | other.name || | +----------------+-------------------------------+-----------------------------+| +Projection | 13 | other.name -- anon[27], other | other.name || | +----------------+-------------------------------+-----------------------------+| +SelectOrSemiApply | 13 | anon[27] -- other | anon[27] || |\ +----------------+-------------------------------+-----------------------------+| | +Expand(All) | 15 | anon[66], anon[80] -- other | (other)-[:WORKS_IN]->() || | | +----------------+-------------------------------+-----------------------------+| | +Argument | 14 | other | || | +----------------+-------------------------------+-----------------------------+| +LetSemiApply | 14 | anon[27] -- other | || |\ +----------------+-------------------------------+-----------------------------+| | +Expand(All) | 2 | anon[35], anon[53] -- other | (other)-[:FRIENDS_WITH]->() || | | +----------------+-------------------------------+-----------------------------+| | +Argument | 14 | other | || | +----------------+-------------------------------+-----------------------------+| +NodeByLabelScan | 14 | other | :Person |+--------------------+----------------+-------------------------------+-----------------------------+


13.3.5. LetAntiSemiApply

Tests for the absence of a pattern predicate. When a query contains multiple pattern predicatesLetAntiSemiApply will be used to evaluate the first of these. It will record the result of evaluating thepredicate but will leave any filtering to another operator. The following query will find all the peoplewho don’t have any friends or who work somewhere.

Query

MATCH (other:Person)WHERE NOT ((other)-[:FRIENDS_WITH]->()) OR (other)-[:WORKS_IN]->()RETURN other.name

Finds all the people who don’t have any friends or who work somewhere. The LetAntiSemiApplyoperator will be used to check for the absence of the FRIENDS_WITH relationship from each person.

193

Query plan

+--------------------+----------------+-------------------------------+-----------------------------+| Operator | Estimated Rows | Variables | Other |+--------------------+----------------+-------------------------------+-----------------------------+| +ProduceResults | 11 | other.name | other.name || | +----------------+-------------------------------+-----------------------------+| +Projection | 11 | other.name -- anon[31], other | other.name || | +----------------+-------------------------------+-----------------------------+| +SelectOrSemiApply | 11 | anon[31] -- other | anon[31] || |\ +----------------+-------------------------------+-----------------------------+| | +Expand(All) | 15 | anon[71], anon[85] -- other | (other)-[:WORKS_IN]->() || | | +----------------+-------------------------------+-----------------------------+| | +Argument | 14 | other | || | +----------------+-------------------------------+-----------------------------+| +LetAntiSemiApply | 14 | anon[31] -- other | || |\ +----------------+-------------------------------+-----------------------------+| | +Expand(All) | 2 | anon[39], anon[57] -- other | (other)-[:FRIENDS_WITH]->() || | | +----------------+-------------------------------+-----------------------------+| | +Argument | 14 | other | || | +----------------+-------------------------------+-----------------------------+| +NodeByLabelScan | 14 | other | :Person |+--------------------+----------------+-------------------------------+-----------------------------+


13.3.6. SelectOrSemiApply

Tests for the existence of a pattern predicate and evaluates a predicate. This operator allows for themixing of normal predicates and pattern predicates that check for the existence of a pattern. First thenormal expression predicate is evaluated, and only if it returns false the costly pattern predicateevaluation is performed.

Query

MATCH (other:Person)WHERE other.age > 25 OR (other)-[:FRIENDS_WITH]->()RETURN other.name

Finds the names of all people who have friends, or are older than 25.

Query plan

+--------------------+----------------+-----------------------------+-----------------------------+| Operator | Estimated Rows | Variables | Other |+--------------------+----------------+-----------------------------+-----------------------------+| +ProduceResults | 11 | other.name | other.name || | +----------------+-----------------------------+-----------------------------+| +Projection | 11 | other.name -- other | other.name || | +----------------+-----------------------------+-----------------------------+| +SelectOrSemiApply | 11 | other | other.age > { AUTOINT0} || |\ +----------------+-----------------------------+-----------------------------+| | +Expand(All) | 2 | anon[53], anon[71] -- other | (other)-[:FRIENDS_WITH]->() || | | +----------------+-----------------------------+-----------------------------+| | +Argument | 14 | other | || | +----------------+-----------------------------+-----------------------------+| +NodeByLabelScan | 14 | other | :Person |+--------------------+----------------+-----------------------------+-----------------------------+


13.3.7. SelectOrAntiSemiApply

Tests for the absence of a pattern predicate and evaluates a predicate.

194

Query

MATCH (other:Person)WHERE other.age > 25 OR NOT (other)-[:FRIENDS_WITH]->()RETURN other.name

Finds the names of all people who do not have friends, or are older than 25.

Query plan

+------------------------+----------------+-----------------------------+-----------------------------+| Operator | Estimated Rows | Variables | Other |+------------------------+----------------+-----------------------------+-----------------------------+| +ProduceResults | 4 | other.name | other.name || | +----------------+-----------------------------+-----------------------------+| +Projection | 4 | other.name -- other | other.name || | +----------------+-----------------------------+-----------------------------+| +SelectOrAntiSemiApply | 4 | other | other.age > { AUTOINT0} || |\ +----------------+-----------------------------+-----------------------------+| | +Expand(All) | 2 | anon[57], anon[75] -- other | (other)-[:FRIENDS_WITH]->() || | | +----------------+-----------------------------+-----------------------------+| | +Argument | 14 | other | || | +----------------+-----------------------------+-----------------------------+| +NodeByLabelScan | 14 | other | :Person |+------------------------+----------------+-----------------------------+-----------------------------+


13.3.8. ConditionalApply

Checks whether a variable is not null, and if so the right-hand side will be executed.

Query

MERGE (p:Person { name: 'Andres' })ON MATCH SET p.exists = TRUE

Looks for the existence of a person called Andres, and if found sets the exists property to true.

Query plan

+-----------------------+----------------+-----------+---------------+| Operator | Estimated Rows | Variables | Other |+-----------------------+----------------+-----------+---------------+| +ProduceResults | 1 | | || | +----------------+-----------+---------------+| +EmptyResult | | | || | +----------------+-----------+---------------+| +AntiConditionalApply | 1 | p | || |\ +----------------+-----------+---------------+| | +MergeCreateNode | 1 | p | || | +----------------+-----------+---------------+| +ConditionalApply | 1 | p | || |\ +----------------+-----------+---------------+| | +SetNodeProperty | 1 | p | || | | +----------------+-----------+---------------+| | +Argument | 1 | p | || | +----------------+-----------+---------------+| +Optional | 1 | p | || | +----------------+-----------+---------------+| +NodeIndexSeek | 1 | p | :Person(name) |+-----------------------+----------------+-----------+---------------+


195

13.3.9. AntiConditionalApply

Checks whether a variable is null, and if so the right-hand side will be executed.

Query

MERGE (p:Person { name: 'Andres' })ON CREATE SET p.exists = TRUE

Looks for the existence of a person called Andres, and if not found, creates one and sets the existsproperty to true.

Query plan

+-----------------------+----------------+-----------+---------------+| Operator | Estimated Rows | Variables | Other |+-----------------------+----------------+-----------+---------------+| +ProduceResults | 1 | | || | +----------------+-----------+---------------+| +EmptyResult | | | || | +----------------+-----------+---------------+| +AntiConditionalApply | 1 | p | || |\ +----------------+-----------+---------------+| | +SetNodeProperty | 1 | p | || | | +----------------+-----------+---------------+| | +MergeCreateNode | 1 | p | || | +----------------+-----------+---------------+| +Optional | 1 | p | || | +----------------+-----------+---------------+| +NodeIndexSeek | 1 | p | :Person(name) |+-----------------------+----------------+-----------+---------------+


13.3.10. AssertSameNode

This operator is used to ensure that no uniqueness constraints are violated.

Query

MERGE (t:Team { name: 'Engineering', id: 42 })

Looks for the existence of a team with the supplied name and id, and if one does not exist, it will becreated. Due to the existence of two uniqueness constraints on :Team(name) and :Team(id), any nodethat would be found by the `UniqueIndexSeek`s must be the very same node, or the constraintswould be violated.

196

Query plan

+---------------------------------+----------------+-----------+-------------+| Operator | Estimated Rows | Variables | Other |+---------------------------------+----------------+-----------+-------------+| +ProduceResults | 1 | | || | +----------------+-----------+-------------+| +EmptyResult | | | || | +----------------+-----------+-------------+| +AntiConditionalApply | 1 | t | || |\ +----------------+-----------+-------------+| | +MergeCreateNode | 1 | t | || | +----------------+-----------+-------------+| +Optional | 1 | t | || | +----------------+-----------+-------------+| +AssertSameNode | 0 | t | || |\ +----------------+-----------+-------------+| | +NodeUniqueIndexSeek(Locking) | 1 | t | :Team(id) || | +----------------+-----------+-------------+| +NodeUniqueIndexSeek(Locking) | 1 | t | :Team(name) |+---------------------------------+----------------+-----------+-------------+


13.3.11. NodeHashJoin

Using a hash table, a NodeHashJoin joins the input coming from the left with the input coming from theright.

Query

MATCH (andy:Person { name:'Andreas' })-[:WORKS_IN]->(loc)<-[:WORKS_IN]-(matt:Person { name:'Mattis' })RETURN loc.name

Returns the name of the location where the matched persons both work.

197

Query plan

+------------------+----------------+-------------------------------------------------+---------------------------+| Operator | Estimated Rows | Variables | Other|+------------------+----------------+-------------------------------------------------+---------------------------+| +ProduceResults | 10 | loc.name | loc.name|| | +----------------+-------------------------------------------------+---------------------------+| +Projection | 10 | loc.name -- anon[37], anon[56], andy, loc, matt | loc.name|| | +----------------+-------------------------------------------------+---------------------------+| +Filter | 10 | anon[37], anon[56], andy, loc, matt | NOT(anon[37] ==anon[56]) || | +----------------+-------------------------------------------------+---------------------------+| +NodeHashJoin | 10 | anon[37], andy -- anon[56], loc, matt | loc|| |\ +----------------+-------------------------------------------------+---------------------------+| | +Expand(All) | 19 | anon[56], loc -- matt | (matt)-[:WORKS_IN]->(loc) || | | +----------------+-------------------------------------------------+---------------------------+| | +NodeIndexSeek | 1 | matt | :Person(name)|| | +----------------+-------------------------------------------------+---------------------------+| +Expand(All) | 19 | anon[37], loc -- andy | (andy)-[:WORKS_IN]->(loc) || | +----------------+-------------------------------------------------+---------------------------+| +NodeIndexSeek | 1 | andy | :Person(name)|+------------------+----------------+-------------------------------------------------+---------------------------+


13.3.12. Triadic

Triadic is used to solve triangular queries, such as the very common 'find my friend-of-friends that arenot already my friend'. It does so by putting all the friends in a set, and use that set to check if thefriend-of-friends are already connected to me.

Query

MATCH (me:Person)-[:FRIENDS_WITH]-()-[:FRIENDS_WITH]-(other)WHERE NOT (me)-[:FRIENDS_WITH]-(other)RETURN other.name

Finds the names of all friends of my friends that are not already my friends.

198

Query plan

+-------------------+----------------+-------------------------------------------------------+----------------------------+| Operator | Estimated Rows | Variables | Other|+-------------------+----------------+-------------------------------------------------------+----------------------------+| +ProduceResults | 0 | other.name | other.name|| | +----------------+-------------------------------------------------------+----------------------------+| +Projection | 0 | other.name -- anon[18], anon[35], anon[37], me, other | other.name|| | +----------------+-------------------------------------------------------+----------------------------+| +TriadicSelection | 0 | anon[18], anon[35], me -- anon[37], other | me,anon[35], other || |\ +----------------+-------------------------------------------------------+----------------------------+| | +Filter | 0 | anon[37], other |NOT(anon[18] == anon[37]) || | | +----------------+-------------------------------------------------------+----------------------------+| | +Expand(All) | 0 | anon[37], other | ()-[:FRIENDS_WITH]-(other) || | | +----------------+-------------------------------------------------------+----------------------------+| | +Argument | 4 | ||| | +----------------+-------------------------------------------------------+----------------------------+| +Expand(All) | 4 | anon[18], anon[35] -- me | (me)-[:FRIENDS_WITH]-() || | +----------------+-------------------------------------------------------+----------------------------+| +NodeByLabelScan | 14 | me | :Person|+-------------------+----------------+-------------------------------------------------------+----------------------------+


13.4. Row operators

These operators take rows produced by another operator and transform them to a different set ofrows

13.4.1. Eager

For isolation purposes this operator makes sure that operations that affect subsequent operations areexecuted fully for the whole dataset before continuing execution. Otherwise it could trigger endlessloops, matching data again, that was just created. The Eager operator can cause high memory usagewhen importing data or migrating graph structures. In such cases split up your operations intosimpler steps e.g. you can import nodes and relationships separately. Alternatively return the recordsto be updated and run an update statement afterwards.

Query

MATCH (a)-[r]-(b)DELETE r,a,bMERGE ()

199

Query Plan

+-------------------------+----------------+------+---------+---------------------+--------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-------------------------+----------------+------+---------+---------------------+--------------+| +ProduceResults | 1 | 0 | 0 | | || | +----------------+------+---------+---------------------+--------------+| +EmptyResult | | 0 | 0 | | || | +----------------+------+---------+---------------------+--------------+| +Apply | 1 | 504 | 0 | a, b, r -- anon[38] | || |\ +----------------+------+---------+---------------------+--------------+| | +AntiConditionalApply | 1 | 504 | 0 | anon[38] | || | |\ +----------------+------+---------+---------------------+--------------+| | | +MergeCreateNode | 1 | 0 | 0 | anon[38] | || | | +----------------+------+---------+---------------------+--------------+| | +Optional | 35 | 504 | 0 | anon[38] | || | | +----------------+------+---------+---------------------+--------------+| | +AllNodesScan | 35 | 504 | 540 | anon[38] | || | +----------------+------+---------+---------------------+--------------+| +Eager | | 36 | 0 | a, b, r | || | +----------------+------+---------+---------------------+--------------+| +Delete(3) | 36 | 36 | 39 | a, b, r | || | +----------------+------+---------+---------------------+--------------+| +Eager | | 36 | 0 | a, b, r | || | +----------------+------+---------+---------------------+--------------+| +Expand(All) | 36 | 36 | 71 | a, r -- b | (b)-[r:]-(a) || | +----------------+------+---------+---------------------+--------------+| +AllNodesScan | 35 | 35 | 36 | b | |+-------------------------+----------------+------+---------+---------------------+--------------+


13.4.2. Distinct

Removes duplicate rows from the incoming stream of rows.

Query

MATCH (l:Location)<-[:WORKS_IN]-(p:Person)RETURN DISTINCT l

Query Plan

+------------------+----------------+------+---------+------------------+----------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+------------------+----------------+------+---------+------------------+----------------------+| +ProduceResults | 14 | 6 | 0 | l | l || | +----------------+------+---------+------------------+----------------------+| +Distinct | 14 | 6 | 0 | l | l || | +----------------+------+---------+------------------+----------------------+| +Filter | 15 | 15 | 15 | anon[19], l, p | p:Person || | +----------------+------+---------+------------------+----------------------+| +Expand(All) | 15 | 15 | 25 | anon[19], p -- l | (l)<-[:WORKS_IN]-(p) || | +----------------+------+---------+------------------+----------------------+| +NodeByLabelScan | 10 | 10 | 11 | l | :Location |+------------------+----------------+------+---------+------------------+----------------------+


13.4.3. Eager Aggregation

Eagerly loads underlying results and stores it in a hash-map, using the grouping keys as the keys forthe map.

Query

MATCH (l:Location)<-[:WORKS_IN]-(p:Person)RETURN l.name AS location, COLLECT(p.name) AS people

200

Query Plan

+-------------------+----------------+------+---------+----------------------------+----------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+-------------------+----------------+------+---------+----------------------------+----------------------+| +ProduceResults | 4 | 6 | 0 | location, people | location, people|| | +----------------+------+---------+----------------------------+----------------------+| +EagerAggregation | 4 | 6 | 15 | people -- location | location|| | +----------------+------+---------+----------------------------+----------------------+| +Projection | 15 | 15 | 15 | location -- anon[19], l, p | l.name; p|| | +----------------+------+---------+----------------------------+----------------------+| +Filter | 15 | 15 | 15 | anon[19], l, p | p:Person|| | +----------------+------+---------+----------------------------+----------------------+| +Expand(All) | 15 | 15 | 25 | anon[19], p -- l | (l)<-[:WORKS_IN]-(p)|| | +----------------+------+---------+----------------------------+----------------------+| +NodeByLabelScan | 10 | 10 | 11 | l | :Location|+-------------------+----------------+------+---------+----------------------------+----------------------+


13.4.4. Node Count From Count Store

Use the count store to answer questions about node counts. This is much faster than eageraggregation which achieves the same result by actually counting. However the count store only savesa limited range of combinations, so eager aggregation will still be used for more complex queries. Forexample, we can get counts for all nodes, and nodes with a label, but not nodes with more than onelabel.

Query

MATCH (p:Person)RETURN count(p) AS people

Query Plan

+--------------------------+----------------+------+---------+-----------+------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+--------------------------+----------------+------+---------+-----------+------------------------------+| +ProduceResults | 4 | 1 | 0 | people | people || | +----------------+------+---------+-----------+------------------------------+| +NodeCountFromCountStore | 4 | 1 | 0 | people | count( (:Person) ) AS people |+--------------------------+----------------+------+---------+-----------+------------------------------+


13.4.5. Relationship Count From Count Store

Use the count store to answer questions about relationship counts. This is much faster than eageraggregation which achieves the same result by actually counting. However the count store only savesa limited range of combinations, so eager aggregation will still be used for more complex queries. Forexample, we can get counts for all relationships, relationships with a type, relationships with a label onone end, but not relationships with labels on both end nodes.

201

Query

MATCH (p:Person)-[r:WORKS_IN]->()RETURN count(r) AS jobs

Query Plan

+----------------------------------+----------------+------+---------+-----------+--------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+----------------------------------+----------------+------+---------+-----------+--------------------------------------------+| +ProduceResults | 4 | 1 | 0 | jobs | jobs|| | +----------------+------+---------+-----------+--------------------------------------------+| +RelationshipCountFromCountStore | 4 | 1 | 1 | jobs | count( (:Person)-[:WORKS_IN]->() ) AS jobs |+----------------------------------+----------------+------+---------+-----------+--------------------------------------------+


13.4.6. Filter

Filters each row coming from the child operator, only passing through rows that evaluate thepredicates to TRUE.

Query

MATCH (p:Person)WHERE p.name =~ "^a.*"RETURN p

Query Plan

+------------------+----------------+------+---------+-----------+-----------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+------------------+----------------+------+---------+-----------+-----------------------------+| +ProduceResults | 14 | 0 | 0 | p | p || | +----------------+------+---------+-----------+-----------------------------+| +Filter | 14 | 0 | 14 | p | p.name ~= /{ AUTOSTRING0}/ || | +----------------+------+---------+-----------+-----------------------------+| +NodeByLabelScan | 14 | 14 | 15 | p | :Person |+------------------+----------------+------+---------+-----------+-----------------------------+


13.4.7. Limit

Returns the first 'n' rows from the incoming input.

Query

MATCH (p:Person)RETURN pLIMIT 3

202

Query Plan

+------------------+----------------+------+---------+-----------+------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+------------------+----------------+------+---------+-----------+------------+| +ProduceResults | 3 | 3 | 0 | p | p || | +----------------+------+---------+-----------+------------+| +Limit | 3 | 3 | 0 | p | Literal(3) || | +----------------+------+---------+-----------+------------+| +NodeByLabelScan | 14 | 3 | 4 | p | :Person |+------------------+----------------+------+---------+-----------+------------+


13.4.8. Projection

For each row from its input, projection evaluates a set of expressions and produces a row with theresults of the expressions.

Query

RETURN "hello" AS greeting

Query Plan

+-----------------+----------------+------+---------+-----------+-----------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+-----------------+| +ProduceResults | 1 | 1 | 0 | greeting | greeting || | +----------------+------+---------+-----------+-----------------+| +Projection | 1 | 1 | 0 | greeting | { AUTOSTRING0} |+-----------------+----------------+------+---------+-----------+-----------------+


13.4.9. Skip

Skips 'n' rows from the incoming rows

Query

MATCH (p:Person)RETURN pORDER BY p.idSKIP 1

203

Query Plan

+------------------+----------------+------+---------+----------------------------+-----------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+------------------+----------------+------+---------+----------------------------+-----------------------+| +ProduceResults | 14 | 13 | 0 | p | p|| | +----------------+------+---------+----------------------------+-----------------------+| +Projection | 14 | 13 | 0 | p -- anon[35], anon[59], p | anon[35]|| | +----------------+------+---------+----------------------------+-----------------------+| +Skip | 14 | 13 | 0 | anon[35], anon[59], p | { AUTOINT0}|| | +----------------+------+---------+----------------------------+-----------------------+| +Sort | 14 | 14 | 0 | anon[35], anon[59], p | anon[59]|| | +----------------+------+---------+----------------------------+-----------------------+| +Projection | 14 | 14 | 28 | anon[59] -- anon[35], p | anon[35]; anon[35].id|| | +----------------+------+---------+----------------------------+-----------------------+| +Projection | 14 | 14 | 0 | anon[35] -- p | p|| | +----------------+------+---------+----------------------------+-----------------------+| +NodeByLabelScan | 14 | 14 | 15 | p | :Person|+------------------+----------------+------+---------+----------------------------+-----------------------+


13.4.10. Sort

Sorts rows by a provided key.

Query

MATCH (p:Person)RETURN pORDER BY p.name

204

Query Plan

+------------------+----------------+------+---------+----------------------------+-------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+------------------+----------------+------+---------+----------------------------+-------------------------+| +ProduceResults | 14 | 14 | 0 | p | p|| | +----------------+------+---------+----------------------------+-------------------------+| +Projection | 14 | 14 | 0 | p -- anon[24], anon[37], p | anon[24]|| | +----------------+------+---------+----------------------------+-------------------------+| +Sort | 14 | 14 | 0 | anon[24], anon[37], p | anon[37]|| | +----------------+------+---------+----------------------------+-------------------------+| +Projection | 14 | 14 | 14 | anon[37] -- anon[24], p | anon[24];anon[24].name || | +----------------+------+---------+----------------------------+-------------------------+| +Projection | 14 | 14 | 0 | anon[24] -- p | p|| | +----------------+------+---------+----------------------------+-------------------------+| +NodeByLabelScan | 14 | 14 | 15 | p | :Person|+------------------+----------------+------+---------+----------------------------+-------------------------+


13.4.11. Top

Returns the first 'n' rows sorted by a provided key. The physical operator is called Top. Instead ofsorting the whole input, only the top X rows are kept.

Query

MATCH (p:Person)RETURN pORDER BY p.nameLIMIT 2

205

Query Plan

+------------------+----------------+------+---------+----------------------------+-------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+------------------+----------------+------+---------+----------------------------+-------------------------+| +ProduceResults | 2 | 2 | 0 | p | p|| | +----------------+------+---------+----------------------------+-------------------------+| +Projection | 2 | 2 | 0 | p -- anon[24], anon[37], p | anon[24]|| | +----------------+------+---------+----------------------------+-------------------------+| +Top | 2 | 2 | 0 | anon[24], anon[37], p | Literal(2); anon[37]|| | +----------------+------+---------+----------------------------+-------------------------+| +Projection | 14 | 14 | 14 | anon[37] -- anon[24], p | anon[24];anon[24].name || | +----------------+------+---------+----------------------------+-------------------------+| +Projection | 14 | 14 | 0 | anon[24] -- p | p|| | +----------------+------+---------+----------------------------+-------------------------+| +NodeByLabelScan | 14 | 14 | 15 | p | :Person|+------------------+----------------+------+---------+----------------------------+-------------------------+


13.4.12. Union

Union concatenates the results from the right plan after the results of the left plan.

Query

MATCH (p:Location)RETURN p.nameUNION ALL MATCH (p:Country)RETURN p.name

Query Plan

+--------------------+----------------+------+---------+-------------+-----------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+--------------------+----------------+------+---------+-------------+-----------+| +ProduceResults | 10 | 11 | 0 | p.name | p.name || | +----------------+------+---------+-------------+-----------+| +Union | 10 | 11 | 0 | p.name | || |\ +----------------+------+---------+-------------+-----------+| | +Projection | 1 | 1 | 1 | p.name -- p | p.name || | | +----------------+------+---------+-------------+-----------+| | +NodeByLabelScan | 1 | 1 | 2 | p | :Country || | +----------------+------+---------+-------------+-----------+| +Projection | 10 | 10 | 10 | p.name -- p | p.name || | +----------------+------+---------+-------------+-----------+| +NodeByLabelScan | 10 | 10 | 11 | p | :Location |+--------------------+----------------+------+---------+-------------+-----------+


13.4.13. Unwind

Takes a list of values and returns one row per item in the list.

206

Query

UNWIND range(1,5) AS valueRETURN value;

Query Plan

+-----------------+----------------+------+---------+-----------+-------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+-------+| +ProduceResults | 10 | 5 | 0 | value | value || | +----------------+------+---------+-----------+-------+| +Unwind | 10 | 5 | 0 | value | || | +----------------+------+---------+-----------+-------+| +EmptyRow | 1 | 1 | 0 | | |+-----------------+----------------+------+---------+-----------+-------+


13.4.14. Call Procedure

Return all labels sorted by name

Query

CALL db.labels() YIELD labelRETURN *ORDER BY label

Query Plan

+-----------------+----------------+------+---------+-----------+----------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other |+-----------------+----------------+------+---------+-----------+----------------------------------+| +ProduceResults | 10000 | 4 | 0 | label | label || | +----------------+------+---------+-----------+----------------------------------+| +Sort | 10000 | 4 | 0 | label | label || | +----------------+------+---------+-----------+----------------------------------+| +ProcedureCall | 10000 | 4 | 1 | label | db.labels() :: (label :: String) || | +----------------+------+---------+-----------+----------------------------------+| +EmptyRow | 1 | 1 | 0 | | |+-----------------+----------------+------+---------+-----------+----------------------------------+


13.5. Update Operators

These operators are used in queries that update the graph.

13.5.1. Constraint Operation

Creates a constraint on a (label,property) pair. The following query will create a unique constraint onthe name property of nodes with the Country label.

Query

CREATE CONSTRAINT ON (c:Country) ASSERT c.name IS UNIQUE

207

Query Plan

+-------------------------+| Operator |+-------------------------+| +CreateUniqueConstraint |+-------------------------+


13.5.2. Empty Result

Eagerly loads everything coming in to the EmptyResult operator and discards it.

Query

CREATE (:Person)

Query Plan

+-----------------+----------------+------+---------+-----------+| Operator | Estimated Rows | Rows | DB Hits | Variables |+-----------------+----------------+------+---------+-----------+| +ProduceResults | 1 | 0 | 0 | || | +----------------+------+---------+-----------+| +EmptyResult | | 0 | 0 | || | +----------------+------+---------+-----------+| +CreateNode | 1 | 1 | 2 | anon[8] |+-----------------+----------------+------+---------+-----------+


13.5.3. Update Graph

Applies updates to the graph.

Query

CYPHER planner=ruleCREATE (:Person { name: "Alistair" })

Query Plan

+--------------+------+---------+-----------+------------+| Operator | Rows | DB Hits | Variables | Other |+--------------+------+---------+-----------+------------+| +EmptyResult | 0 | 0 | | || | +------+---------+-----------+------------+| +UpdateGraph | 1 | 4 | anon[7] | CreateNode |+--------------+------+---------+-----------+------------+


13.5.4. Merge Into

When both the start and end node have already been found, merge-into is used to find all connectingrelationships or creating a new relationship between the two nodes.

208

Query

CYPHER planner=ruleMATCH (p:Person { name: "me" }),(f:Person { name: "Andres" })MERGE (p)-[:FRIENDS_WITH]->(f)

Query Plan

+--------------+------+---------+------------------+--------------------------------+| Operator | Rows | DB Hits | Variables | Other |+--------------+------+---------+------------------+--------------------------------+| +EmptyResult | 0 | 0 | | || | +------+---------+------------------+--------------------------------+| +Merge(Into) | 1 | 5 | anon[68] -- f, p | (p)-[:FRIENDS_WITH]->(f) || | +------+---------+------------------+--------------------------------+| +SchemaIndex | 1 | 2 | f -- p | { AUTOSTRING1}; :Person(name) || | +------+---------+------------------+--------------------------------+| +SchemaIndex | 1 | 2 | p | { AUTOSTRING0}; :Person(name) |+--------------+------+---------+------------------+--------------------------------+


13.6. Shortest path planning

Shortest path finding in Cypher and how it is planned.

Planning shortest paths in Cypher can lead to different query plans depending on the predicates thatneed to be evaluated. Internally, Neo4j will use a fast bidirectional breadth-first search algorithm if thepredicates can be evaluated whilst searching for the path. Therefore, this fast algorithm will always becertain to return the right answer when there are universal predicates on the path; for example, whensearching for the shortest path where all nodes have the Person label, or where there are no nodeswith a name property.

If the predicates need to inspect the whole path before deciding on whether it is valid or not, this fastalgorithm cannot be relied on to find the shortest path, and Neo4j may have to resort to using aslower exhaustive depth-first search algorithm to find the path. This means that query plans forshortest path queries with non-universal predicates will include a fallback to running the exhaustivesearch to find the path should the fast algorithm not succeed. For example, depending on the data, ananswer to a shortest path query with existential predicates — such as the requirement that at leastone node contains the property name='Charlie Sheen' — may not be able to be found by the fastalgorithm. In this case, Neo4j will fall back to using the exhaustive search to enumerate all paths andpotentially return an answer.

The running times of these two algorithms may differ by orders of magnitude, so it is important toensure that the fast approach is used for time-critical queries.

When the exhaustive search is planned, it is still only executed when the fast algorithm fails to findany matching paths. The fast algorithm is always executed first, since it is possible that it can find avalid path even though that could not be guaranteed at planning time.

13.6.1. Shortest path with fast algorithm

Query

MATCH (ms:Person { name:'Martin Sheen' }),(cs:Person { name:'Charlie Sheen' }), p = shortestPath((ms)-[rels:ACTED_IN*]-(cs))WHERE ALL (r IN rels WHERE exists(r.role))RETURN p

This query can be evaluated with the fast algorithm — there are no predicates that need to see the

209

whole path before being evaluated.

Query plan

+--------------------+----------------+------+---------+-------------------+--------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+--------------------+----------------+------+---------+-------------------+--------------------------------------+| +ProduceResults | 1 | 1 | 0 | p | p|| | +----------------+------+---------+-------------------+--------------------------------------+| +ShortestPath | 1 | 1 | 9 | p, rels -- cs, ms | all(r in rels wherehasProp(r.role)) || | +----------------+------+---------+-------------------+--------------------------------------+| +CartesianProduct | 1 | 1 | 0 | ms -- cs ||| |\ +----------------+------+---------+-------------------+--------------------------------------+| | +Filter | 1 | 1 | 5 | cs | cs.name == { AUTOSTRING1}|| | | +----------------+------+---------+-------------------+--------------------------------------+| | +NodeByLabelScan | 5 | 5 | 6 | cs | :Person|| | +----------------+------+---------+-------------------+--------------------------------------+| +Filter | 1 | 1 | 5 | ms | ms.name == { AUTOSTRING0}|| | +----------------+------+---------+-------------------+--------------------------------------+| +NodeByLabelScan | 5 | 5 | 6 | ms | :Person|+--------------------+----------------+------+---------+-------------------+--------------------------------------+


13.6.2. Shortest path with additional predicate checks on the paths

Consider using the exhaustive search as a fallback

Predicates used in the WHERE clause that apply to the shortest path pattern are evaluated beforedeciding what the shortest matching path is.

Query

MATCH (cs:Person { name:'Charlie Sheen' }),(ms:Person { name:'Martin Sheen' }), p = shortestPath((cs)-[*]-(ms))WHERE length(p)> 1RETURN p

This query, in contrast with the one above, needs to check that the whole path follows the predicatebefore we know if it is valid or not, and so the query plan will also include the fallback to the slowerexhaustive search algorithm

210

Query plan

+--------------------------+----------------+------+---------+----------------------------------+-----------------------------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+--------------------------+----------------+------+---------+----------------------------------+-----------------------------------------------+| +ProduceResults | 0 | 1 | 0 | p | p|| | +----------------+------+---------+----------------------------------+-----------------------------------------------+| +AntiConditionalApply | 0 | 1 | 0 | cs -- anon[93], anon[112], ms, p ||| |\ +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Top1 | 0 | 0 | 0 | anon[93], anon[112], ms, p | anon[93]|| | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Projection | 0 | 0 | 0 | anon[93] -- anon[112], ms, p |length(p) || | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Filter | 0 | 0 | 0 | anon[112], ms, p |length(p) > { AUTOINT2} || | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Projection | 0 | 0 | 0 | p -- anon[112], ms |ProjectedPath(Set(anon[112], cs),<function2>) || | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +VarLengthExpand(Into) | 0 | 0 | 0 | anon[112], ms | (cs)-[:*]->(ms) || | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Argument | 1 | 0 | 0 | ||| | +----------------+------+---------+----------------------------------+-----------------------------------------------+| +Apply | 1 | 1 | 0 | cs, ms -- anon[112], p ||| |\ +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Optional | 1 | 1 | 0 | anon[112], p ||| | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +ShortestPath | 0 | 1 | 1 | anon[112], p |length(p) > { AUTOINT2} || | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Argument | 1 | 1 | 0 | ||| | +----------------+------+---------+----------------------------------+-----------------------------------------------+| +CartesianProduct | 1 | 1 | 0 | cs -- ms ||| |\ +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +Filter | 1 | 1 | 5 | ms | ms.name== { AUTOSTRING1} || | | +----------------+------+---------+----------------------------------+-----------------------------------------------+| | +NodeByLabelScan | 5 | 5 | 6 | ms | :Person|| | +----------------+------+---------+----------------------------------+-----------------------------------------------+| +Filter | 1 | 1 | 5 | cs | cs.name== { AUTOSTRING0} || | +----------------+------+---------+----------------------------------+-----------------------------------------------+| +NodeByLabelScan | 5 | 5 | 6 | cs | :Person|+--------------------------+----------------+------+---------+----------------------------------+-----------------------------------------------+


211

The way the bigger exhaustive query plan works is by using Apply/Optional to ensure that when thefast algorithm does not find any results, a NULL result is generated instead of simply stopping theresult stream. On top of this, the planner will issue an AntiConditionalApply, which will run theexhaustive search if the path variable is pointing to NULL instead of a path.

Prevent the exhaustive search from being used as a fallback

Query

MATCH (cs:Person { name:'Charlie Sheen' }),(ms:Person { name:'Martin Sheen' }), p = shortestPath((cs)-[*]-(ms))WITH pWHERE length(p)> 1RETURN p

This query, just like the one above, needs to check that the whole path follows the predicate before weknow if it is valid or not. However, the inclusion of the WITH clause means that the query plan will notinclude the fallback to the slower exhaustive search algorithm. Instead, any paths found by the fastalgorithm will subsequently be filtered, which may result in no answers being returned.

Query plan

+--------------------+----------------+------+---------+-----------------------------------+-----------------------------+| Operator | Estimated Rows | Rows | DB Hits | Variables | Other|+--------------------+----------------+------+---------+-----------------------------------+-----------------------------+| +ProduceResults | 1 | 1 | 0 | p | p|| | +----------------+------+---------+-----------------------------------+-----------------------------+| +Filter | 1 | 1 | 0 | anon[146], anon[112], cs, ms, p | anon[146]|| | +----------------+------+---------+-----------------------------------+-----------------------------+| +Projection | 1 | 1 | 0 | anon[146] -- anon[112], cs, ms, p | p; length(p)> { AUTOINT2} || | +----------------+------+---------+-----------------------------------+-----------------------------+| +ShortestPath | 1 | 1 | 1 | anon[112], p -- cs, ms ||| | +----------------+------+---------+-----------------------------------+-----------------------------+| +CartesianProduct | 1 | 1 | 0 | cs -- ms ||| |\ +----------------+------+---------+-----------------------------------+-----------------------------+| | +Filter | 1 | 1 | 5 | ms | ms.name == {AUTOSTRING1} || | | +----------------+------+---------+-----------------------------------+-----------------------------+| | +NodeByLabelScan | 5 | 5 | 6 | ms | :Person|| | +----------------+------+---------+-----------------------------------+-----------------------------+| +Filter | 1 | 1 | 5 | cs | cs.name == {AUTOSTRING0} || | +----------------+------+---------+-----------------------------------+-----------------------------+| +NodeByLabelScan | 5 | 5 | 6 | cs | :Person|+--------------------+----------------+------+---------+-----------------------------------+-----------------------------+


212

DriversNeo4j Drivers are drivers for the Neo4j database (http://neo4j.com/). The drivers use the Bolt protocoland have uniform design and use.

213

http://neo4j.com/

Chapter 14. IntroductionThe preferred way to access a Neo4j server from an application is to use a Neo4j driver.

The Neo4j Driver API is the preferred means of programmatic interaction with a Neo4j databaseserver. It implements the Bolt protocol and is available in four languages: C#.NET, Java, JavaScript, andPython.

The API is defined independently of any programming language. This allows for a high degree ofuniformity across languages. Uniformity means that the same features are included in all the drivers.The uniformity also influences the design of the API in each language. This provides consistencyacross drivers, while retaining affinity with the idioms of each programming language.

214

Chapter 15. Getting StartedGet up and running quickly with a minimal working example.

15.1. Get the driver

15.1.1. Versions

Consult the version table to determine which version of the driver to use with a particular Neo4jserver.

Table 49. Supported versions

Driver version Neo4j version Bolt protocol version

1.0 3.0 1

You can download the driver source or acquire it with one of the dependency managers of yourlanguage.

Example 1. Example: Acquire the driver

npm install [email protected]

15.2. Use the driver

Each Neo4j driver has a database object for creating a driver. To use a driver, follow this pattern:

1. Ask the database object for a new driver.

2. Ask the driver object for a new session.

3. Use the session object to run statements. It returns an object representing the results.

4. Process the results.

5. Close the session.

215

Example 2. Example: Use the driver

For a minimal working example, the following imports are necessary.

var neo4j = require('neo4j-driver').v1;

These can then be used to run a statement against the server.

var driver = neo4j.driver("bolt://localhost", neo4j.auth.basic("neo4j", "neo4j"));var session = driver.session();session .run( "CREATE (a:Person {name:'Arthur', title:'King'})" ) .then( function() { return session.run( "MATCH (a:Person) WHERE a.name = 'Arthur' RETURN a.name AS name, a.title AStitle" ) }) .then( function( result ) { console.log( result.records[0].get("title") + " " + result.records[0].get("name") ); session.close(); driver.close(); })

As seen above, it is possible to run statements directly from the session object. The session will takecare of opening and closing the transaction. The session also provides the opportunity for a user tomanage the transaction explicitly. See Transaction management for details.

216

Chapter 16. DriverA driver is used to connect to a Neo4j server. It provides sessions that are used to execute

statements and retrieve results.

16.1. Construction

The driver package includes a graph database object. This object provides driver instances. Whenrequesting a driver instance from the database object a URL is provided. The URL declares theprotocol, host name, and port for the Neo4j server.

16.1.1. Bolt URL format

A Bolt URL follows the standard URL pattern of scheme://hostname:port. For example,bolt://server:1234 would connect using the Bolt protocol to server on port 1234. If no port isprovided in the URL, the default port 7687 is used. For example, bolt://server amounts to the sameas bolt://server:7687.

Example 3. Example: Create a driver

driverGlobal = neo4j.driver("bolt://localhost", neo4j.auth.basic("neo4j", "neo4j"));

16.2. Configuration

In addition to the Bolt URL, the driver can be configured for:

• session pool size

• session pool behavior

• authentication strategies

• logging

Name Description Default

Session pool size (*) Max number of sessions per URL. 50

Logging Provide a logging facility for the driver. N/A

(*) The .NET and Python drivers provide idle session pool size, which is the maximum number of idlesessions to be pooled by the driver. There is no limit to how many sessions that can be created, but amaximum limits how many sessions will be buffered after they are returned to the session pool.

Example 4. Example: Create a driver with configuration

var driver = neo4j.driver("bolt://localhost", neo4j.auth.basic("neo4j", "neo4j"), {connectionPoolSize: 50});

16.2.1. Encryption and authentication

All Neo4j drivers support encrypting the traffic between the Neo4j driver and the Neo4j instance using

217

SSL/TLS. Neo4j will by default allow both encrypted and unencrypted connections from drivers. Thiscan be modified to require encryption for all connections. Please see the Neo4j Operations Manual →Configuring Bolt Connectors (http://neo4j.com/docs/operations-manual/3.0/#_configuring_bolt_connectors) formore information.

We strongly encourage using encryption to keep authentication credentials and data stored in Neo4jsecure. Because of this most drivers will turn encryption on by default. However, due to technicallimitations, the .NET driver does not have encryption enabled by default. For similar reasons, theJavaScript driver does not have encryption enabled by default when running inside a Web Browser.

Encryption

The driver can be configured to turn on or off TLS encryption. Encryption should only be turned off ina trusted, internal network. While most drivers default to enabling encryption, it is good practice toconfigure encryption explicitly. This makes it clear encryption is used for other developers reading thecode and minimizes the risk of mistakes from relying on defaults.

Example 5. Example: Configure driver to require TLS encryption

var driver = neo4j.driver("bolt://localhost", neo4j.auth.basic("neo4j", "neo4j"), { // In NodeJS, encryption is on by default. In the web bundle, it is off. encrypted:true});

Trust

When establishing an encrypted connection, it needs to be verified that the remote peer is who weexpected to connect to. Without verifying this, an attacker can easily perform a so-called "man-in-the-middle" attack, tricking the driver to establish an encrypted connection with her instead of the Neo4jInstance. Because of this, Neo4j drivers do not offer a way to establish encrypted connections withoutalso establishing trust.

Two trust strategies are available:

• Trust on first use.

• Trust signed certificate.

Trust on first use means that the driver will trust the first connection to a host to be safe andintentional. On subsequent connections, the driver will verify that it communicates with the same hostas on the first connection. To ensure that this authentication strategy is valid, the first connection tothe server must be established in a trusted network environment.

This strategy is the same as the default strategy used by the ssh command line tool. It provides a gooddegree of security with no configuration. For this reason it is the default strategy on all platforms thatsupport it.

Because of technical limitations this strategy is not available for the .NET platform or when using theJavaScript driver inside a web browser.

218

http://neo4j.com/docs/operations-manual/3.0/#_configuring_bolt_connectors

http://neo4j.com/docs/operations-manual/3.0/#_configuring_bolt_connectors

Example 6. Example: Configure driver to trust on first use

var driver = neo4j.driver("bolt://localhost", neo4j.auth.basic("neo4j", "neo4j"), { // Note that trust-on-first-use is not available in the browser bundle, // in NodeJS, trust-on-first-use is the default trust mode. In the browser // it is TRUST_SIGNED_CERTIFICATES. trust: "TRUST_ON_FIRST_USE", encrypted:true});

Trust signed certificate means that the driver will only connect to a Neo4j server if it provides acertificate trusted by the driver. A trusted certificate is a certificate explicitly configured to be trustedby the driver or, more commonly, a certificate signed by a certificate the driver trusts. All driverssupport this trust strategy.

The way to install a trusted certificate for use by a driver varies for different drivers. The Java driverdirectly accepts the file path to the trusted certificate and loads the certificate automatically atruntime. The .NET and Python drivers require the certificate to first be installed into the operatingsystem’s trusted certificate store.

Example 7. Example: Configure driver to trust signed certificate

var driver = neo4j.driver("bolt://localhost", neo4j.auth.basic("neo4j", "neo4j"), { trust: "TRUST_SIGNED_CERTIFICATES", // Configuring which certificates to trust here is only available // in NodeJS. In the browser bundle the browsers list of trusted // certificates is used, due to technical limitations in some browsers. trustedCertificates : ["path/to/ca.crt"], encrypted:true});

Authenticate user

The server will require the driver to provide authentication credentials for the user to connect to thedatabase, unless authentication has been disabled.

Example 8. Example: Connecting the driver to a server with authentication disabled

var driver = neo4j.driver("bolt://localhost", { // In NodeJS, encryption is on by default. In the web bundle, it is off. encrypted:true});

If communication between the driver and the server is not encrypted, authentication credentials aresent as plain text. It is highly recommended to use encryption when using authentication.

When connecting to Neo4j for the first time, the user must change the defaultpassword. Given a fresh Neo4j server installation, any attempt to connect to theNeo4j server with the default password will provide a notifcation indicating that thecredentials have expired and the password needs to be updated. The browser canbe used to change the expired password.

219

16.3. Lifecycle

For most use cases it is recommended to use a single driver instance throughout an application. Theonly reason to create multiple driver instances is to connect to multiple servers. This may be usefulwhen connecting to multiple members of a cluster, or for pulling data out of one database to push toanother. In other cases, a single driver instance usually serves the needs of a single application.

220

Chapter 17. SessionAll interaction with a Neo4j server takes place within a session.

A session is used to run statements against Neo4j. A session is obtained from a driver object. It allowsfor running statements against the database.

The driver has a session pool which can be configured. Sessions are returned to the pool when theyare closed. It is important to close sessions, to allow them to return to the pool and be reused.

It is important to provide for the closing of a session in the case where an exceptionis thrown. Make sure that the exceptional path of execution, as well as the ordinary,closes the session.

17.1. Run a statement

The session can run statements directly. If a statement is run directly from a session, the session willtake care of wrapping it in a transaction.

Example 9. Example: Run a statement

session .run( "CREATE (person:Person {name: {name}})", {name: "Arthur"} )

It is always recommended to use parameters, but it is possible to run a query with literal values aswell.

Why it’s important to use parameters

Using parameters have significant performance and security benefits, including:

• It allows the query planner to reuse its plans, which makes queries much moreefficient.

• It protects the database from Cypher injection attacks, where malicious queryclauses are added from poorly typed or filtered input.

Example 10. Example: Run a statement without parameters

session .run( "CREATE (p:Person { name: 'Arthur' })" )

17.2. Transaction management

When the session runs a statement directly, it creates an implicit transaction. The session can alsocreate explicit transactions to be managed manually. When explicitly told to begin a new transaction,the session will return a transaction object. The transaction object allows for fine-grained transactioncontrol.

For details on transactions in Neo4j see the Neo4j Java Developer Reference(http://neo4j.com/docs/java-reference/3.0/#transactions).

221

http://neo4j.com/docs/java-reference/3.0/#transactions

Example 11. Example: Run a statement in a manually managed transaction

var tx = session.beginTransaction();tx.run( "CREATE (:Person {name: 'Guinevere'})" );tx.commit();

The manually managed transaction can also be rolled back.

Example 12. Example: Roll back a manually managed transaction

var tx = session.beginTransaction();tx.run( "CREATE (:Person {name: 'Merlin'})" );tx.rollback();

222

Chapter 18. ResultsResults returned from Neo4j are presented as a stream of records.

When a statement is run, results are returned as a record stream. The stream itself is represented by aresult cursor which is used to access the individual records. The individual records can be accessedone at a time, or all at once. Results are valid until the next statement is run or until the end of thecurrent transaction, whichever comes first. To hold on to the results while running additionalstatements, or to use the results outside of the scope of the current transaction, the results must beretained by assigning them to a variable.

18.1. The result cursor

A result cursor provides access to the stream of records that make up the results. The cursor is movedthrough the stream and points to each record in turn. Before the cursor is moved onto the firstrecord, it points before the first record. When the cursor has traversed to the end of the result stream,summary information and metadata are available.

Result records are loaded lazily as the cursor is moved through the stream. Thismeans that the cursor must be moved to the first result before this result can beconsumed. It also means that the entire stream must be explicitly consumed beforesummary information and metadata is available. It is generally best practice toexplicitly consume results and close sessions, in particular when running updatestatements. This applies even if the summary information is not required. Failing toconsume a result can lead to unpredictable behaviour as there will be no guaranteethat the server has seen and handled the Cypher statement.

Example 13. Example: Use a result cursor

var searchTerm = "Sword";session .run( "MATCH (weapon:Weapon) WHERE weapon.name CONTAINS {term} RETURN weapon.name", {term :searchTerm} ) .subscribe({ onNext: function(record) { console.log("" + record.get("weapon.name")); }, onCompleted: function() { session.close(); }, onError: function(error) { console.log(error); } });

18.2. The record

The result record stream is made up of individual records. A record provides an immutable view of apart of a result. It is an ordered map of keys and values. These key-value pairs are called fields. As thefields are both keyed and ordered, the value of a field can be accessed either by position (0-basedInteger) or by key (String).

Access values by key or position

• To access values by position, use a 0-based integer.

• To access values by key, use a string.

223

Example 14. Example: Use a record

var searchTerm = "Arthur";session .run( "MATCH (weapon:Weapon) WHERE weapon.owner CONTAINS {term} RETURN weapon.name,weapon.material, weapon.size", {term : searchTerm} ) .subscribe({ onNext: function(record) { var sword = []; record.forEach(function(value, key) { sword.push(key + ": " + value); }); console.log(sword); }, onCompleted: function() { session.close(); }, onError: function(error) { console.log(error); } });

18.2.1. Retaining Results

The result record stream is available until another statement is run in the session, or until the currenttransaction is closed. To hold on to the results beyond this scope, the results need to be explicitlyretained. For retaining results, each driver offers methods that collect the result stream and translate itinto standard data structures for each language. Retained results can be processed, while the sessionis free to take on the next workload. The saved results can also be used directly to run a newstatement.

Example 15. Example: Retain results for further processing

session .run("MATCH (knight:Person:Knight) WHERE knight.castle = {castle} RETURN knight.name AS name",{castle: "Camelot"}) .then(function (result) { var records = []; for (i = 0; i < result.records.length; i++) { records.push(result.records[i]); } return records; }) .then(function (records) { for(i = 0; i < records.length; i ++) { console.log(records[i].get("name") + " is a knight of Camelot"); } });

224

Example 16. Example: Use results in a new query

session .run("MATCH (knight:Person:Knight) WHERE knight.castle = {castle} RETURN id(knight) AS knight_id",{"castle": "Camelot"}) .subscribe({ onNext: function(record) { session .run("MATCH (knight) WHERE id(knight) = {id} MATCH (king:Person) WHERE king.name = {king}CREATE (knight)-[:DEFENDS]->(king)", {"id": record.get("knight_id"), "king": "Arthur"}); }, onCompleted: function() { session .run("MATCH (:Knight)-[:DEFENDS]->() RETURN count(*)") .then(function (result) { console.log("Count is " + result.records[0].get(0).toInt()); }); }, onError: function(error) { console.log(error); } });

18.3. Result summary

Metadata can be retrieved along with the result stream. An example is the query plan obtained byprepending a Cypher query with PROFILE or EXPLAIN. The following example will produce an explainplan with accurate numbers by running the query in the database. This will output a query profile (seethe Neo4j Manual (http://neo4j.com/docs/stable/how-do-i-profile-a-query.html) for more information aboutquery profiling).

Example 17. Example: Profile query and print execution plan

session .run("PROFILE MATCH (p:Person {name: {name}}) RETURN id(p)", {name: "Arthur"}) .then(function (result) { console.log(result.summary.profile); });

Neo4j may also provide notifications for a query, for example, warning when a query will cause theexecution engine to build a cartesian product. The following example will produce an explain plan withestimated numbers, without actually running the query in the database.

Example 18. Example: Explain query and print notifications

session .run("EXPLAIN MATCH (king), (queen) RETURN king, queen") .then(function (result) { var notifications = result.summary.notifications, i; for (i = 0; i < notifications.length; i++) { console.log(notifications[i].code); } });

225

http://neo4j.com/docs/stable/how-do-i-profile-a-query.html

Chapter 19. TypesA driver translates between the type system of its language and the type system used in

Neo4j.

Neo4j has a system of data types for processing and storing data. Drivers translate betweenapplication language types and the Neo4j type system. To pass parameters and process results, it isimportant to know the basics of the Neo4j type system and to understand how the Neo4j types aremapped in the driver.

The following Neo4j types are valid both for parameters and results:

• Boolean

• Integer

• Float

• String

• List

• Map

Maps and lists can be processed by the database and used in a statement. Lists in which the elementsshare a single primitive Neo4j type can be stored as properties on nodes and relationships. Maps, andlists in which the elements have mixed types, can not be stored as properties.

In addition to the types valid both for parameters and results, the following Neo4j types are valid forresults only:

• Node

• Relationship

• Path

The Neo4j driver maps Neo4j types to native language types as follows:

Example 19. Map Neo4j types to native language types

Neo4j JavaScript

Null null

Boolean Boolean

Integer Integer (*)

Float Number

String String

List Array

Map Object

Node Node (*)

Relationship Relationship (*)

Path Path (*)

226

Chapter 20. ErrorsThere are two types of errors that the user of a driver may need to be aware of. The more likely errorto encounter is the Cypher error. A Cypher error occurs when the server is unable to run thestatement that it receives from the driver.

Cypher errors encountered using a Neo4j driver are the same as those that can beencountered anywhere Cypher is used. This includes the Neo4j Browser, Neo4jShell, the Cypher HTTP endpoint, the embedded API, and within a Java storedprocedure. See Neo4j Status Codes for further details.

Example 20. Handling a Cypher error

session .run("Then will cause a syntax error") .catch( function(err) { expect(err.fields[0].code).toBe( "Neo.ClientError.Statement.SyntaxError" ); done(); });

HTTP APIThe HTTP API.

227

Chapter 21. Transactional Cypher HTTPendpointThe Neo4j transactional HTTP endpoint allows you to execute a series of Cypher statements within thescope of a transaction. The transaction may be kept open across multiple HTTP requests, until theclient chooses to commit or roll back. Each HTTP request can include a list of statements, and forconvenience you can include statements along with a request to begin or commit a transaction.

The server guards against orphaned transactions by using a timeout. If there are no requests for agiven transaction within the timeout period, the server will roll it back. You can configure the timeoutin the server configuration, by setting dbms.transaction_timeout to the number of seconds beforetimeout. The default timeout is 60 seconds.

• Literal line breaks are not allowed inside Cypher statements.

• Open transactions are not shared among members of an HA cluster. Therefore,if you use this endpoint in an HA cluster, you must ensure that all requests for agiven transaction are sent to the same Neo4j instance.

• Cypher queries with USING PERIODIC COMMIT (see Using Periodic Commit)may only be executed when creating a new transaction and immediatelycommitting it with a single HTTP request (see Begin and commit a transaction inone request for how to do that).

• When a request fails the transaction will be rolled back. By checking the resultfor the presence/absence of the transaction key you can figure out if thetransaction is still open.

21.1. Streaming

Responses from the HTTP API can be transmitted as JSON streams, resulting in better performanceand lower memory overhead on the server side. To use streaming, supply the header X-Stream: truewith each request.

In order to speed up queries in repeated scenarios, try not to use literals butreplace them with parameters wherever possible. This will let the server cachequery plans. See Parameters for more information.

21.2. Begin and commit a transaction in one request

If there is no need to keep a transaction open across multiple HTTP requests, you can begin atransaction, execute statements, and commit with just a single HTTP request.

Example request

• POST http://localhost:7474/db/data/transaction/commit

• Accept: application/json; charset=UTF-8

• Content-Type: application/json

{ "statements" : [ { "statement" : "CREATE (n) RETURN id(n)" } ]}

Example response

228

• 200: OK


{ "results" : [ { "columns" : [ "id(n)" ], "data" : [ { "row" : [ 78 ], "meta" : [ null ] } ] } ], "errors" : [ ]}

21.3. Execute multiple statements

You can send multiple Cypher statements in the same request. The response will contain the result ofeach statement.

Example request




{ "statements" : [ { "statement" : "CREATE (n) RETURN id(n)" }, { "statement" : "CREATE (n {props}) RETURN n", "parameters" : { "props" : { "name" : "My Node" } } } ]}

Example response

• 200: OK


229

{ "results" : [ { "columns" : [ "id(n)" ], "data" : [ { "row" : [ 74 ], "meta" : [ null ] } ] }, { "columns" : [ "n" ], "data" : [ { "row" : [ { "name" : "My Node" } ], "meta" : [ { "id" : 75, "type" : "node", "deleted" : false } ] } ] } ], "errors" : [ ]}

21.4. Begin a transaction

You begin a new transaction by posting zero or more Cypher statements to the transaction endpoint.The server will respond with the result of your statements, as well as the location of your opentransaction.

Example request

• POST http://localhost:7474/db/data/transaction



{ "statements" : [ { "statement" : "CREATE (n {props}) RETURN n", "parameters" : { "props" : { "name" : "My Node" } } } ]}

Example response

• 201: Created


• Location: http://localhost:7474/db/data/transaction/46

230

{ "commit" : "http://localhost:7474/db/data/transaction/46/commit", "results" : [ { "columns" : [ "n" ], "data" : [ { "row" : [ { "name" : "My Node" } ], "meta" : [ { "id" : 82, "type" : "node", "deleted" : false } ] } ] } ], "transaction" : { "expires" : "Fri, 06 May 2016 00:59:43 +0000" }, "errors" : [ ]}

21.5. Execute statements in an open transaction

Given that you have an open transaction, you can make a number of requests, each of which executesadditional statements, and keeps the transaction open by resetting the transaction timeout.

Example request

• POST http://localhost:7474/db/data/transaction/48



{ "statements" : [ { "statement" : "CREATE (n) RETURN n" } ]}

Example response

• 200: OK


{ "commit" : "http://localhost:7474/db/data/transaction/48/commit", "results" : [ { "columns" : [ "n" ], "data" : [ { "row" : [ { } ], "meta" : [ { "id" : 83, "type" : "node", "deleted" : false } ] } ] } ], "transaction" : { "expires" : "Fri, 06 May 2016 00:59:43 +0000" }, "errors" : [ ]}

231

21.6. Reset transaction timeout of an open transaction

Every orphaned transaction is automatically expired after a period of inactivity. This may beprevented by resetting the transaction timeout.

The timeout may be reset by sending a keep-alive request to the server that executes an empty list ofstatements. This request will reset the transaction timeout and return the new time at which thetransaction will expire as an RFC1123 formatted timestamp value in the ``transaction'' section of theresponse.

Example request




{ "statements" : [ ]}

Example response

• 200: OK


{ "commit" : "http://localhost:7474/db/data/transaction/38/commit", "results" : [ ], "transaction" : { "expires" : "Fri, 06 May 2016 00:59:42 +0000" }, "errors" : [ ]}

21.7. Commit an open transaction

Given you have an open transaction, you can send a commit request. Optionally, you submitadditional statements along with the request that will be executed before committing the transaction.

Example request

• POST http://localhost:7474/db/data/transaction/42/commit



{ "statements" : [ { "statement" : "CREATE (n) RETURN id(n)" } ]}

Example response

• 200: OK


232

{ "results" : [ { "columns" : [ "id(n)" ], "data" : [ { "row" : [ 77 ], "meta" : [ null ] } ] } ], "errors" : [ ]}

21.8. Rollback an open transaction

Given that you have an open transaction, you can send a rollback request. The server will rollback thetransaction. Any further statements trying to run in this transaction will fail immediately.

Example request

• DELETE http://localhost:7474/db/data/transaction/39


Example response

• 200: OK

• Content-Type: application/json; charset=UTF-8

{ "results" : [ ], "errors" : [ ]}

21.9. Include query statistics

By setting includeStats to true for a statement, query statistics will be returned for it.

Example request




{ "statements" : [ { "statement" : "CREATE (n) RETURN id(n)", "includeStats" : true } ]}

Example response

• 200: OK


233

{ "results" : [ { "columns" : [ "id(n)" ], "data" : [ { "row" : [ 76 ], "meta" : [ null ] } ], "stats" : { "contains_updates" : true, "nodes_created" : 1, "nodes_deleted" : 0, "properties_set" : 0, "relationships_created" : 0, "relationship_deleted" : 0, "labels_added" : 0, "labels_removed" : 0, "indexes_added" : 0, "indexes_removed" : 0, "constraints_added" : 0, "constraints_removed" : 0 } } ], "errors" : [ ]}

21.10. Return results in graph format

If you want to understand the graph structure of nodes and relationships returned by your query, youcan specify the "graph" results data format. For example, this is useful when you want to visualise thegraph structure. The format collates all the nodes and relationships from all columns of the result, andalso flattens collections of nodes and relationships, including paths.

Example request




{ "statements" : [ { "statement" : "CREATE ( bike:Bike { weight: 10 } ) CREATE ( frontWheel:Wheel { spokes: 3 } ) CREATE (backWheel:Wheel { spokes: 32 } ) CREATE p1 = (bike)-[:HAS { position: 1 } ]->(frontWheel) CREATE p2 =(bike)-[:HAS { position: 2 } ]->(backWheel) RETURN bike, p1, p2", "resultDataContents" : [ "row", "graph" ] } ]}

Example response

• 200: OK


{ "results" : [ { "columns" : [ "bike", "p1", "p2" ], "data" : [ { "row" : [ { "weight" : 10 }, [ { "weight" : 10 }, { "position" : 1 }, { "spokes" : 3 } ], [ { "weight" : 10

234

}, { "position" : 2 }, { "spokes" : 32 } ] ], "meta" : [ { "id" : 79, "type" : "node", "deleted" : false }, [ { "id" : 79, "type" : "node", "deleted" : false }, { "id" : 33, "type" : "relationship", "deleted" : false }, { "id" : 80, "type" : "node", "deleted" : false } ], [ { "id" : 79, "type" : "node", "deleted" : false }, { "id" : 34, "type" : "relationship", "deleted" : false }, { "id" : 81, "type" : "node", "deleted" : false } ] ], "graph" : { "nodes" : [ { "id" : "80", "labels" : [ "Wheel" ], "properties" : { "spokes" : 3 } }, { "id" : "81", "labels" : [ "Wheel" ], "properties" : { "spokes" : 32 } }, { "id" : "79", "labels" : [ "Bike" ], "properties" : { "weight" : 10 } } ], "relationships" : [ { "id" : "33", "type" : "HAS", "startNode" : "79", "endNode" : "80", "properties" : { "position" : 1 } }, { "id" : "34", "type" : "HAS", "startNode" : "79", "endNode" : "81", "properties" : { "position" : 2 } } ] } } ] } ], "errors" : [ ]}

235

21.11. Handling errors

The result of any request against the transaction endpoint is streamed back to the client. Thereforethe server does not know whether the request will be successful or not when it sends the HTTP statuscode.

Because of this, all requests against the transactional endpoint will return 200 or 201 status code,regardless of whether statements were successfully executed. At the end of the response payload, theserver includes a list of errors that occurred while executing statements. If this list is empty, therequest completed successfully.

If any errors occur while executing statements, the server will roll back the transaction.

In this example, we send the server an invalid statement to demonstrate error handling.

For more information on the status codes, see Neo4j Status Codes.

Example request

• POST http://localhost:7474/db/data/transaction/47/commit



{ "statements" : [ { "statement" : "This is not a valid Cypher Statement." } ]}

Example response

• 200: OK


{ "results" : [ ], "errors" : [ { "code" : "Neo.ClientError.Statement.SyntaxError", "message" : "Invalid input 'T': expected <init> (line 1, column 1 (offset: 0))\n\"This is not a validCypher Statement.\"\n ^" } ]}

21.12. Handling errors in an open transaction

Whenever there is an error in a request the server will rollback the transaction. By inspecting theresponse for the presence/absence of the transaction key you can tell if the transaction is still open

Example request




236

{ "statements" : [ { "statement" : "This is not a valid Cypher Statement." } ]}

Example response

• 200: OK


{ "commit" : "http://localhost:7474/db/data/transaction/45/commit", "results" : [ ], "errors" : [ { "code" : "Neo.ClientError.Statement.SyntaxError", "message" : "Invalid input 'T': expected <init> (line 1, column 1 (offset: 0))\n\"This is not a validCypher Statement.\"\n ^" } ]}

237

Chapter 22. Authentication and AuthorizationIn order to prevent unauthorized access to Neo4j, the HTTP API supports authorization andauthentication. When enabled, requests to the HTTP API must be authorized using the username andpassword of a valid user. Authorization is enabled by default, see the Operations Manual(http://neo4j.com/docs/operations-manual/3.0/#security-server-auth) for how to disable it.

When Neo4j is first installed you can authenticate with the default user neo4j and the defaultpassword neo4j. However, the default password must be changed (see User status and passwordchanging) before access to resources will be permitted. This can easily be done via the Neo4j Browser,or via direct HTTP calls.

The username and password combination is local to each Neo4j instance. If you wish to have multipleinstances in a cluster, you should ensure that all instances share the same credential. For automateddeployments, you may also copy security configuration from another Neo4j instance (see Copyingsecurity configuration from one instance to another).

22.1. Authenticating

22.1.1. Missing authorization

If an Authorization header is not supplied, the server will reply with an error.

Example request

• GET http://localhost:7474/db/data/


Example response

• 401: Unauthorized


• WWW-Authenticate: Basic realm="Neo4j"

{ "errors" : [ { "code" : "Neo.ClientError.Security.Unauthorized", "message" : "No authentication header supplied." } ]}

22.1.2. Authenticate to access the server

Authenticate by sending a username and a password to Neo4j using HTTP Basic Auth. Requestsshould include an Authorization header, with a value of Basic <payload>, where "payload" is a base64encoded string of "username:password".

Example request

• GET http://localhost:7474/user/neo4j


• Authorization: Basic bmVvNGo6c2VjcmV0

Example response

238

http://neo4j.com/docs/operations-manual/3.0/#security-server-auth

• 200: OK


{ "password_change_required" : false, "password_change" : "http://localhost:7474/user/neo4j/password", "username" : "neo4j"}

22.1.3. Incorrect authentication

If an incorrect username or password is provided, the server replies with an error.

Example request

• POST http://localhost:7474/db/data/


• Authorization: Basic bmVvNGo6aW5jb3JyZWN0

Example response

• 401: Unauthorized


• WWW-Authenticate: Basic realm="Neo4j"

{ "errors" : [ { "code" : "Neo.ClientError.Security.Unauthorized", "message" : "Invalid username or password." } ]}

22.1.4. Required password changes

In some cases, like the very first time Neo4j is accessed, the user will be required to choose a newpassword. The database will signal that a new password is required and deny access.

See [rest-api-security-user-status-and-password-changing] for how to set a new password.

Example request

• GET http://localhost:7474/db/data/


• Authorization: Basic bmVvNGo6bmVvNGo=

Example response

• 403: Forbidden


239

{ "password_change" : "http://localhost:7474/user/neo4j/password", "errors" : [ { "code" : "Neo.ClientError.Security.Forbidden", "message" : "User is required to change their password." } ]}

22.2. User status and password changing

22.2.1. User status

Given that you know the current password, you can ask the server for the user status.

Example request



• Authorization: Basic bmVvNGo6c2VjcmV0

Example response

• 200: OK


{ "password_change_required" : false, "password_change" : "http://localhost:7474/user/neo4j/password", "username" : "neo4j"}

22.2.2. User status on first access

On first access, and using the default password, the user status will indicate that the users passwordrequires changing.

Example request




Example response

• 200: OK


{ "password_change_required" : true, "password_change" : "http://localhost:7474/user/neo4j/password", "username" : "neo4j"}

240

22.2.3. Changing the user password

Given that you know the current password, you can ask the server to change a users password. Youcan choose any password you like, as long as it is different from the current password.

Example request

• POST http://localhost:7474/user/neo4j/password




{ "password" : "secret"}

Example response

• 200: OK

22.3. Access when auth is disabled

When auth has been disabled in the configuration, requests can be sent without an Authorizationheader.

22.4. Copying security configuration from one instance toanother

In many cases, such as automated deployments, you may want to start a Neo4j instance with pre-configured authentication and authorization. This is possible by copying the auth database file from apre-existing Neo4j instance to your new instance.

This file is located at data/dbms/auth, and simply copying that file into a new Neo4j instance willtransfer your password and authorization token.

ProceduresUser-defined procedures are written in Java, deployed into the database, and called from

Cypher.

A procedure is a mechanism that allows Neo4j to be extended by writing custom code which can beinvoked directly from Cypher. Procedures can take arguments, perform operations on the database,and return results.

Procedures are written in Java and compiled into jar files. They can be deployed to the database bydropping a jar file into the $NEO4J_HOME/plugins directory on each standalone or clustered server. Thedatabase must be re-started on each server to pick up new procedures.

Procedures are the preferred means for extending Neo4j. Examples of use cases for procedures are:

241

1. To provide access to functionality that is not available in Cypher, such as manual indexes andschema introspection.

2. To provide access to third party systems.

3. To perform graph-global operations, such as counting connected components or finding densenodes.

4. To express a procedural operation that is difficult to express declaratively with Cypher.

242

Chapter 23. Calling proceduresTo call a stored procedure, use a Cypher CALL clause. The procedure name must be fully qualified, so aprocedure named findDenseNodes defined in the package org.neo4j.examples could be called using:

CALL org.neo4j.examples.findDenseNodes(1000)

A CALL may be the only clause within a Cypher statement or may be combined with other clauses.Arguments can be supplied directly within the query or taken from the associated parameter set. Forfull details, see the Cypher documentation on the CALL clause.

243

Chapter 24. Built-in proceduresNeo4j comes bundled with a number of built-in procedures. These are listed in the table below:

Procedure name Command to invoke procedure What it does

ListLabels CALL db.labels() List all labels in the database.

ListRelationshipTypes CALL db.relationshipTypes() List all relationship types in thedatabase.

ListPropertyKeys CALL db.propertyKeys() List all property keys in the database.

ListIndexes CALL db.indexes() List all indexes in the database.

ListConstraints CALL db.constraints() List all constraints in the database.

ListProcedures CALL dbms.procedures() List all procedures in the DBMS.

ListComponents CALL dbms.components() List DBMS components and theirversions.

QueryJmx CALL dbms.queryJmx(query) Query JMX management data bydomain and name. For instance,"org.neo4j:*".

AlterUserPassword CALL dbms.changePassword(query) Change the user password.

244

Chapter 25. User-defined procedures

The example discussed below is available as a repository on GitHub(https://github.com/neo4j-examples/neo4j-procedure-template). To get started quickly youcan fork the repository and work with the code as you follow along in the guidebelow.

Custom procedures are written in the Java programming language. Procedures are deployed via a jarfile that contains the code itself along with any dependencies (excluding Neo4j). These files should beplaced into the plugin directory of each standalone database or cluster member and will becomeavailable following the next database restart.

The example that follows shows the steps to create and deploy a new procedure.

25.1. Set up a new project

A project can be set up in any way that allows for compiling a procedure and producing a jar file.Below is an example configuration using the Maven (https://maven.apache.org/) build system. Forreadability, only excerpts from the Maven pom.xml file are shown here, the whole file is available fromthe Neo4j Procedure Template (https://github.com/neo4j-examples/neo4j-procedure-template) repository.

Setting up a project with Maven

<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd"> <modelVersion>4.0.0</modelVersion>

<groupId>org.neo4j.example</groupId> <artifactId>procedure-template</artifactId> <version>1.0.0-SNAPSHOT</version>

<packaging>jar</packaging> <name>Neo4j Procedure Template</name> <description>A template project for building a Neo4j Procedure</description>

<properties> <neo4j.version>3.0.0-SNAPSHOT</neo4j.version> </properties>

Next, the build dependencies are defined. The following two sections are included in the pom.xmlbetween <dependencies></dependencies> tags.

The first dependency section includes the procedure API that procedures use at runtime. The scope isset to provided, because once the procedure is deployed to a Neo4j instance, this dependency isprovided by Neo4j. If non-Neo4j dependencies are added to the project, their scope should normallybe compile.

<dependency> <groupId>org.neo4j</groupId> <artifactId>neo4j</artifactId> <version>${neo4j.version}</version> <scope>provided</scope></dependency>

Next, the dependencies necessary for testing the procedure are added:

• Neo4j Harness, a utility that allows for starting a lightweight Neo4j instance. It is used to startNeo4j with a specific procedure deployed, which greatly simplifies testing.

• The Neo4j Java driver, used to send cypher statements that call the procedure.

245

https://github.com/neo4j-examples/neo4j-procedure-template

https://maven.apache.org/


• JUnit, a common Java test framework.

<dependency> <groupId>org.neo4j.test</groupId> <artifactId>neo4j-harness</artifactId> <version>${neo4j.version}</version> <scope>test</scope></dependency>

<dependency> <groupId>org.neo4j.driver</groupId> <artifactId>neo4j-java-driver</artifactId> <version>1.0-SNAPSHOT</version> <scope>test</scope></dependency>

<dependency> <groupId>junit</groupId> <artifactId>junit</artifactId> <version>4.12</version> <scope>test</scope></dependency>

Along with declaring the dependencies used by the procedure it is also necessary to define the stepsthat Maven will go through to build the project. The goal is first to compile the source, then to packageit in a jar that can be deployed to a Neo4j instance.

Procedures require at least Java 8, so the version 1.8 should be defined as thesource and target version in the configuration for the Maven compiler plugin.

The Maven Shade (https://maven.apache.org/plugins/maven-shade-plugin/) plugin is used to package thecompiled procedure. It also includes all dependencies in the package, unless the dependency scope isset to test or provided.

Once the procedure has been deployed to the plugins directory of each Neo4j instance and theinstances have restarted, the procedure is available for use.

<build> <plugins> <plugin> <artifactId>maven-compiler-plugin</artifactId> <configuration> <source>1.8</source> <target>1.8</target> </configuration> </plugin> <plugin> <artifactId>maven-shade-plugin</artifactId> <executions> <execution> <phase>package</phase> <goals> <goal>shade</goal> </goals> </execution> </executions> </plugin> </plugins></build>

Until the GA release of Neo4j 3.0, the dependency on Neo4j requires that a snapshot repository isconfigured. This repository is where Maven will find the latest build of Neo4j to use as a dependency.

246

https://maven.apache.org/plugins/maven-shade-plugin/

<repositories> <repository> <id>neo4j-snapshot-repository</id> <name>Maven 2 snapshot repository for Neo4j</name> <url>http://m2.neo4j.org/content/repositories/snapshots</url> <snapshots><enabled>true</enabled></snapshots> <releases><enabled>false</enabled></releases> </repository></repositories>

25.2. Writing integration tests

The test dependencies include Neo4j Harness and JUnit. These can be used to write integration testsfor procedures.

First, we decide what the procedure should do, then we write a test that proves that it does it right.Finally we write a procedure that passes the test.

Below is a template for testing a procedure that accesses Neo4j’s full-text indexes from Cypher.

Writing tests for procedures

package example;

import org.junit.Rule;import org.junit.Test;import org.neo4j.driver.v1.*;import org.neo4j.graphdb.factory.GraphDatabaseSettings;import org.neo4j.harness.junit.Neo4jRule;

import static org.hamcrest.core.IsEqual.equalTo;import static org.junit.Assert.assertThat;import static org.neo4j.driver.v1.Values.parameters;

public class ManualFullTextIndexTest{ // This rule starts a Neo4j instance for us @Rule public Neo4jRule neo4j = new Neo4jRule()

// This is the Procedure we want to test .withProcedure( FullTextIndex.class );

@Test public void shouldAllowIndexingAndFindingANode() throws Throwable { // In a try-block, to make sure we close the driver after the test try( Driver driver = GraphDatabase.driver( neo4j.boltURI() , Config.build().withEncryptionLevel(Config.EncryptionLevel.NONE ).toConfig() ) ) {

// Given I've started Neo4j with the FullTextIndex procedure class // which my 'neo4j' rule above does. Session session = driver.session();

// And given I have a node in the database long nodeId = session.run( "CREATE (p:User {name:'Brookreson'}) RETURN id(p)" ) .single() .get( 0 ).asLong();

// When I use the index procedure to index a node session.run( "CALL example.index({id}, ['name'])", parameters( "id", nodeId ) );

// Then I can search for that node with lucene query syntax StatementResult result = session.run( "CALL example.search('User', 'name:Brook*')" ); assertThat( result.single().get( "nodeId" ).asLong(), equalTo( nodeId ) ); } }}

247

25.3. Writing a procedure

With the test in place, we write a procedure procedure that fulfils the expectations of the test. The fullexample is available in the Neo4j Procedure Template (https://github.com/neo4j-examples/neo4j-procedure-

template) repository.

Particular things to note:

• All procedures are annotated @Procedure. Procedures that write to the database are additionallyannotated @PerformsWrites.

• The context of the procedure, which is the same as each resource that the procedure wants to use,is annotated @Context.

• The input and output.

For more details, see the API documentation for procedures (http://neo4j.com/docs/java-

reference/3.0/javadocs/index.html?org/neo4j/procedure/Procedure.html).

The correct way to signal an error from within a procedure is to throw aRuntimeException.

package example;

import java.util.List;import java.util.Map;import java.util.Set;import java.util.stream.Stream;

import org.neo4j.graphdb.GraphDatabaseService;import org.neo4j.graphdb.Label;import org.neo4j.graphdb.Node;import org.neo4j.graphdb.index.Index;import org.neo4j.graphdb.index.IndexManager;import org.neo4j.logging.Log;import org.neo4j.procedure.Context;import org.neo4j.procedure.Name;import org.neo4j.procedure.PerformsWrites;import org.neo4j.procedure.Procedure;

import static org.neo4j.helpers.collection.MapUtil.stringMap;

/** * This is an example showing how you could expose Neo4j's full text indexes as * two procedures - one for updating indexes, and one for querying by label and * the lucene query language. */public class FullTextIndex{ // Only static fields and @Context-annotated fields are allowed in // Procedure classes. This static field is the configuration we use // to create full-text indexes. private static final Map<String,String> FULL_TEXT = stringMap( IndexManager.PROVIDER, "lucene", "type", "fulltext" );

// This field declares that we need a GraphDatabaseService // as context when any procedure in this class is invoked @Context public GraphDatabaseService db;

// This gives us a log instance that outputs messages to the // standard log, `neo4j.log` @Context public Log log;

/** * This declares the first of two procedures in this class - a * procedure that performs queries in a manual index. * * It returns a Stream of Records, where records are * specified per procedure. This particular procedure returns * a stream of {@link SearchHit} records.

248


http://neo4j.com/docs/java-reference/3.0/javadocs/index.html?org/neo4j/procedure/Procedure.html

* * The arguments to this procedure are annotated with the * {@link Name} annotation and define the position, name * and type of arguments required to invoke this procedure. * There is a limited set of types you can use for arguments, * these are as follows: * * <ul> * <li>{@link String}</li> * <li>{@link Long} or {@code long}</li> * <li>{@link Double} or {@code double}</li> * <li>{@link Number}</li> * <li>{@link Boolean} or {@code boolean}</li> * <li>{@link java.util.Map} with key {@link String} and value {@link Object}</li> * <li>{@link java.util.List} of elements of any valid argument type, including {@linkjava.util.List}</li> * <li>{@link Object}, meaning any of the valid argument types</li> * </ul> * * @param label the label name to query by * @param query the lucene query, for instance `name:Brook*` to * search by property `name` and find any value starting * with `Brook`. Please refer to the Lucene Query Parser * documentation for full available syntax. * @return the nodes found by the query */ @Procedure("example.search") @PerformsWrites public Stream<SearchHit> search( @Name("label") String label, @Name("query") String query ) { String index = indexName( label );

// Avoid creating the index, if it's not there we won't be // finding anything anyway! if( !db.index().existsForNodes( index )) { // Just to show how you'd do logging log.debug( "Skipping index query since index does not exist: `%s`", index ); return Stream.empty(); }

// If there is an index, do a lookup and convert the result // to our output record. return db.index() .forNodes( index ) .query( query ) .stream() .map( SearchHit::new ); }

/** * This is the second procedure defined in this class, it is used to update the * index with nodes that should be queryable. You can send the same node multiple * times, if it already exists in the index the index will be updated to match * the current state of the node. * * This procedure works largely the same as {@link #search(String, String)}, * with two notable differences. One, it is annotated with {@link PerformsWrites}, * which is <i>required</i> if you want to perform updates to the graph in your * procedure. * * Two, it returns {@code void} rather than a stream. This is simply a short-hand * for saying our procedure always returns an empty stream of empty records. * * @param nodeId the id of the node to index * @param propKeys a list of property keys to index, only the ones the node * actually contains will be added */ @Procedure("example.index") @PerformsWrites public void index( @Name("nodeId") long nodeId, @Name("properties") List<String> propKeys ) { Node node = db.getNodeById( nodeId );

// Load all properties for the node once and in bulk, // the resulting set will only contain those properties in `propKeys` // that the node actually contains. Set<Map.Entry<String,Object>> properties =

249

node.getProperties( propKeys.toArray( new String[0] ) ).entrySet();

// Index every label (this is just as an example, we could filter which labels to index) for ( Label label : node.getLabels() ) { Index<Node> index = db.index().forNodes( indexName( label.name() ), FULL_TEXT );

// In case the node is indexed before, remove all occurrences of it so // we don't get old or duplicated data index.remove( node );

// And then index all the properties for ( Map.Entry<String,Object> property : properties ) { index.add( node, property.getKey(), property.getValue() ); } } }

/** * This is the output record for our search procedure. All procedures * that return results return them as a Stream of Records, where the * records are defined like this one - customized to fit what the procedure * is returning. * * These classes can only have public non-final fields, and the fields must * be one of the following types: * * <ul> * <li>{@link String}</li> * <li>{@link Long} or {@code long}</li> * <li>{@link Double} or {@code double}</li> * <li>{@link Number}</li> * <li>{@link Boolean} or {@code boolean}</li> * <li>{@link org.neo4j.graphdb.Node}</li> * <li>{@link org.neo4j.graphdb.Relationship}</li> * <li>{@link org.neo4j.graphdb.Path}</li> * <li>{@link java.util.Map} with key {@link String} and value {@link Object}</li> * <li>{@link java.util.List} of elements of any valid field type, including {@linkjava.util.List}</li> * <li>{@link Object}, meaning any of the valid field types</li> * </ul> */ public static class SearchHit { // This records contain a single field named 'nodeId' public long nodeId;

public SearchHit( Node node ) { this.nodeId = node.getId(); } }

private String indexName( String label ) { return "label-" + label; }}

Appendix

250

Chapter 26. Neo4j Status CodesThe transactional endpoint may in any response include zero or more status codes, indicating issuesor information for the client. Each status code follows the same format:"Neo.[Classification].[Category].[Title]". The fact that a status code is returned by the server doesalways mean there is a fatal error. Status codes can also indicate transient problems that may go awayif you retry the request.

What the effect of the status code is can be determined by its classification.

This is not the same thing as HTTP status codes. Neo4j Status Codes are returned inthe response body, at the very end of the response.

26.1. Classifications

Classification Description Effect on transaction

ClientError The Client sent a bad request - changing the request might yield asuccessful outcome.

Rollback

ClientNotification There are notifications about the request sent by the client. None

TransientError The database cannot service the request right now, retrying later mightyield a successful outcome.

Rollback

DatabaseError The database failed to service the request. Rollback

26.2. Status codes

This is a complete list of all status codes Neo4j may return, and what they mean.

Status Code Description

Neo.ClientError.General.ForbiddenOnReadOnlyDatabase This is a read only database, writing or modifying thedatabase is not allowed.

Neo.ClientError.LegacyIndex.LegacyIndexNotFound The request (directly or indirectly) referred to a legacyindex that does not exist.

Neo.ClientError.Procedure.ProcedureCallFailed Failed to invoke a procedure. See the detailed errordescription for exact cause.

Neo.ClientError.Procedure.ProcedureNotFound A request referred to a procedure that is not registeredwith this database instance. If you are deploying customprocedures in a cluster setup, ensure all instances in thecluster have the procedure jar file deployed.

Neo.ClientError.Procedure.ProcedureRegistrationFailed The database failed to register a procedure, refer to theassociated error message for details.

Neo.ClientError.Procedure.TypeError A procedure is using or receiving a value of an invalid type.

Neo.ClientError.Request.Invalid The client provided an invalid request.

Neo.ClientError.Request.InvalidFormat The client provided a request that was missing requiredfields, or had values that are not allowed.

Neo.ClientError.Request.TransactionRequired The request cannot be performed outside of a transaction,and there is no transaction present to use. Wrap yourrequest in a transaction and retry.

Neo.ClientError.Schema.ConstraintAlreadyExists Unable to perform operation because it would clash with apre-existing constraint.

Neo.ClientError.Schema.ConstraintNotFound The request (directly or indirectly) referred to a constraintthat does not exist.

Neo.ClientError.Schema.ConstraintValidationFailed A constraint imposed by the database was violated.

251


Neo.ClientError.Schema.ConstraintVerificationFailed Unable to create constraint because data that exists in thedatabase violates it.

Neo.ClientError.Schema.ForbiddenOnConstraintIndex A requested operation can not be performed on thespecified index because the index is part of a constraint. Ifyou want to drop the index, for instance, you must dropthe constraint.

Neo.ClientError.Schema.IndexAlreadyExists Unable to perform operation because it would clash with apre-existing index.

Neo.ClientError.Schema.IndexNotFound The request (directly or indirectly) referred to an index thatdoes not exist.

Neo.ClientError.Schema.TokenNameError A token name, such as a label, relationship type orproperty key, used is not valid. Tokens cannot be emptystrings and cannot be null.

Neo.ClientError.Security.AuthenticationRateLimit The client has provided incorrect authentication details toomany times in a row.

Neo.ClientError.Security.CredentialsExpired The credentials have expired and need to be updated.

Neo.ClientError.Security.EncryptionRequired A TLS encrypted connection is required.

Neo.ClientError.Security.Forbidden An attempt was made to perform an unauthorized action.

Neo.ClientError.Security.Unauthorized The client is unauthorized due to authentication failure.

Neo.ClientError.Statement.ArgumentError The statement is attempting to perform operations usinginvalid arguments

Neo.ClientError.Statement.ArithmeticError Invalid use of arithmetic, such as dividing by zero.

Neo.ClientError.Statement.ConstraintVerificationFailed A constraint imposed by the statement is violated by thedata in the database.

Neo.ClientError.Statement.EntityNotFound The statement is directly referring to an entity that doesnot exist.

Neo.ClientError.Statement.LabelNotFound The statement is referring to a label that does not exist.

Neo.ClientError.Statement.ParameterMissing The statement is referring to a parameter that was notprovided in the request.

Neo.ClientError.Statement.PropertyNotFound The statement is referring to a property that does notexist.

Neo.ClientError.Statement.SemanticError The statement is syntactically valid, but expressessomething that the database cannot do.

Neo.ClientError.Statement.SyntaxError The statement contains invalid or unsupported syntax.

Neo.ClientError.Statement.TypeError The statement is attempting to perform operations onvalues with types that are not supported by the operation.

Neo.ClientError.Transaction.ForbiddenDueToTransactionType

The transaction is of the wrong type to service the request.For instance, a transaction that has had schemamodifications performed in it cannot be used tosubsequently perform data operations, and vice versa.

Neo.ClientError.Transaction.TransactionAccessedConcurrently

There were concurrent requests accessing the sametransaction, which is not allowed.

Neo.ClientError.Transaction.TransactionEventHandlerFailed

A transaction event handler threw an exception. Thetransaction will be rolled back.

Neo.ClientError.Transaction.TransactionHookFailed Transaction hook failure.

Neo.ClientError.Transaction.TransactionMarkedAsFailed Transaction was marked as both successful and failed.Failure takes precedence and so this transaction was rolledback although it may have looked like it was going to becommitted

Neo.ClientError.Transaction.TransactionNotFound The request referred to a transaction that does not exist.

252


Neo.ClientError.Transaction.TransactionTerminated The current transaction has been marked as terminated,meaning no more interactions with it are allowed. Thereare several reasons this happens - the client might haveasked for the transaction to be terminated, an operatormight have asked for the database to be shut down, or thecurrent instance is about to go through a cluster roleswitch. Simply retry your operation in a new transaction.

Neo.ClientError.Transaction.TransactionValidationFailed

Transaction changes did not pass validation checks

Neo.ClientNotification.Statement.CartesianProductWarning

This query builds a cartesian product betweendisconnected patterns.

Neo.ClientNotification.Statement.DynamicPropertyWarning

Queries using dynamic properties will use neither indexseeks nor index scans for those properties

Neo.ClientNotification.Statement.EagerOperatorWarning The execution plan for this query contains the Eageroperator, which forces all dependent data to bematerialized in main memory before proceeding

Neo.ClientNotification.Statement.ExhaustiveShortestPathWarning

Exhaustive shortest path has been planned for your querythat means that shortest path graph algorithm might notbe used to find the shortest path. Hence an exhaustiveenumeration of all paths might be used in order to find therequested shortest path.

Neo.ClientNotification.Statement.FeatureDeprecationWarning

This feature is deprecated and will be removed in futureversions.

Neo.ClientNotification.Statement.JoinHintUnfulfillableWarning

The database was unable to plan a hinted join.

Neo.ClientNotification.Statement.JoinHintUnsupportedWarning

Queries with join hints are not supported by the RULEplanner.

Neo.ClientNotification.Statement.NoApplicableIndexWarning

Adding a schema index may speed up this query.

Neo.ClientNotification.Statement.PlannerUnsupportedWarning

This query is not supported by the COST planner.

Neo.ClientNotification.Statement.RuntimeUnsupportedWarning

This query is not supported by the compiled runtime.

Neo.ClientNotification.Statement.UnboundedVariableLengthPatternWarning

The provided pattern is unbounded, consider adding anupper limit to the number of node hops.

Neo.ClientNotification.Statement.UnknownLabelWarning The provided label is not in the database.

Neo.ClientNotification.Statement.UnknownPropertyKeyWarning

The provided property key is not in the database

Neo.ClientNotification.Statement.UnknownRelationshipTypeWarning

The provided relationship type is not in the database.

Neo.DatabaseError.General.IndexCorruptionDetected The request (directly or indirectly) referred to an index thatis in a failed state. The index needs to be dropped andrecreated manually.

Neo.DatabaseError.General.SchemaCorruptionDetected A malformed schema rule was encountered. Pleasecontact your support representative.

Neo.DatabaseError.General.UnknownError An unknown error occurred.

Neo.DatabaseError.Schema.ConstraintCreationFailed Creating a requested constraint failed.

Neo.DatabaseError.Schema.ConstraintDropFailed The database failed to drop a requested constraint.

Neo.DatabaseError.Schema.IndexCreationFailed Failed to create an index.

Neo.DatabaseError.Schema.IndexDropFailed The database failed to drop a requested index.

Neo.DatabaseError.Schema.LabelAccessFailed The request accessed a label that did not exist.

Neo.DatabaseError.Schema.LabelLimitReached The maximum number of labels supported has beenreached, no more labels can be created.

253


Neo.DatabaseError.Schema.PropertyKeyAccessFailed The request accessed a property that does not exist.

Neo.DatabaseError.Schema.RelationshipTypeAccessFailed The request accessed a relationship type that does notexist.

Neo.DatabaseError.Schema.SchemaRuleAccessFailed The request referred to a schema rule that does not exist.

Neo.DatabaseError.Schema.SchemaRuleDuplicateFound The request referred to a schema rule that is definedmultiple times.

Neo.DatabaseError.Statement.ExecutionFailed The database was unable to execute the statement.

Neo.DatabaseError.Transaction.TransactionCommitFailed The database was unable to commit the transaction.

Neo.DatabaseError.Transaction.TransactionLogError The database was unable to write transaction to log.

Neo.DatabaseError.Transaction.TransactionRollbackFailed

The database was unable to roll back the transaction.

Neo.DatabaseError.Transaction.TransactionStartFailed The database was unable to start the transaction.

Neo.TransientError.General.DatabaseUnavailable The database is not currently available to serve yourrequest, refer to the database logs for more details.Retrying your request at a later time may succeed.

Neo.TransientError.General.OutOfMemoryError There is not enough memory to perform the current task.Please try increasing 'dbms.memory.heap.max_size' in theprocess wrapper configuration (normally in 'conf/neo4j-wrapper.conf' or, if you are using Neo4j Desktop, foundthrough the user interface) or if you are running anembedded installation increase the heap by using '-Xmx'command line flag, and then restart the database.

Neo.TransientError.General.StackOverFlowError There is not enough stack size to perform the current task.This is generally considered to be a database error, soplease contact Neo4j support. You could try increasing thestack size: for example to set the stack size to 2M, add`dbms.jvm.additional=-Xss2M' to in the process wrapperconfiguration (normally in 'conf/neo4j-wrapper.conf' or, ifyou are using Neo4j Desktop, found through the userinterface) or if you are running an embedded installationjust add -Xss2M as command line flag.

Neo.TransientError.Network.CommunicationError An unknown network failure occurred, a retry may resolvethe issue.

Neo.TransientError.Schema.SchemaModifiedConcurrently The database schema was modified while this transactionwas running, the transaction should be retried.

Neo.TransientError.Security.ModifiedConcurrently The user was modified concurrently to this request.

Neo.TransientError.Statement.ExternalResourceFailed Access to an external resource failed

Neo.TransientError.Transaction.ConstraintsChanged Database constraints changed since the start of thistransaction

Neo.TransientError.Transaction.DeadlockDetected This transaction, and at least one more transaction, hasacquired locks in a way that it will wait indefinitely, and thedatabase has aborted it. Retrying this transaction will mostlikely be successful.

Neo.TransientError.Transaction.InstanceStateChanged Transactions rely on assumptions around the state of theNeo4j instance they execute on. For instance, transactionsin a cluster may expect that they are executing on aninstance that can perform writes. However, instances maychange state while the transaction is running. This causesassumptions the instance has made about how to executethe transaction to be violated - meaning the transactionmust be rolled back. If you see this error, you should retryyour operation in a new transaction.

Neo.TransientError.Transaction.LockSessionExpired The lock session under which this transaction was startedis no longer valid.

254

Chapter 27. TerminologyThe terminology used for Cypher and Neo4j is drawn from the worlds of database design and graphtheory. This section provides cross-linked summaries of common terms.

In some cases, multiple terms (e.g., arc, edge, relationship) may be used for the same or similarconcept. An asterisk (*) to the right of a term indicates that the term is commonly used for Neo4j andCypher.

acyclic

for a graph or subgraph: when there is no way to start at some node n and follow a sequence ofadjacent relationships that eventually loops back to n again. The opposite of cyclic.

adjacent

nodes sharing an incident (that is, directly-connected) relationship or relationships sharing anincident node.

attribute

Synonym for property.

arc

graph theory: a synonym for a directed relationship.

array

container that holds a number of elements. The element types can be the types supported by theunderlying graph storage layer, but all elements must be of the same type.

aggregating expression

expression that summarizes a set of values, like computing their sum or their maximum.

clause

component of a Cypher query or command; starts with an identifying keyword (for exampleCREATE). The following clauses currently exist in Cypher: CREATE, CREATE UNIQUE, DELETE, FOREACH, LOADCSV, MATCH, MERGE, OPTIONAL MATCH, REMOVE, RETURN, SET, START, UNION, and WITH.

co-incident

alternative term for adjacent relationships, which share a common node.

collection

container that holds a number of values. The values can have mixed types.

command

a statement that operates on the database without affecting the data graph or returning contentfrom it.

commit

successful completion of a transaction, ensuring durability of any changes made.

constraint

part of a database schema: defines a contract that the database will never break (for example,uniqueness of a property on all nodes that have a specific label).

cyclic

The opposite of acyclic.

255

Cypher

a special-purpose programming language for describing queries and operations on a graphdatabase, with accompanying natural language concepts.

DAG

a directed, acyclic graph: there are no cyclic paths and all the relationships are directed.

data graph

graph stored in the database. See also property graph.

data record

a unit of storage containing an arbitrary unordered collection of properties.

degree

of a node: is the number of relationships leaving or entering (if directed) the node; loops arecounted twice.

directed relationship

a relationship that has a direction; that is the relationship has a source node and a destinationnode. The opposite of an undirected relationship. All relationships in a Neo4j graph are directed.

edge

graph theory: a synonym for undirected relationship.

execution result

all statements return an execution result. For queries, this can contain an iterator of result rows.

execution plan

parsed and compiled statement that is ready for Neo4j to execute. An execution plan consists ofthe physical operations that need to be performed in order to achieve the intent of the statement.

expression

produces values; may be used in projections, as a predicate, or when setting properties on graphelements.

graph

1. data graph,

2. property graph,

3. graph theory: set of vertices and edges.

graph database

a database that uses graph-based structures (for example, nodes, relationships, properties) torepresent and store data.

graph element

a node, relationship, or path which is part of a graph.

variable

variables are named bindings to values (for example, collections, scalars) in a statement. Forexample, in MATCH (n) RETURN n, n is a variable.

incident

adjacent relationship attached to a node or a node attached to a relationship.

incoming relationship

256

pertaining to a directed relationship: from the point of view of a node n, this is any relationship rarriving at n, exemplified by ()-[:r]→(n). The opposite of outgoing.

index

data structure that improves performance of a database by redundantly storing the sameinformation in a way that is faster to read.

intermediate result

set of variables and values (record) passed from one clause to another during query execution.This is internal to the execution of a given query.

label

marks a node as a member of a named subset. A node may be assigned zero or more labels.Labels are written as :label in Cypher (the actual label is prefixed by a colon). Note: graph theory:This differs from mathematical graphs, where a label applies uniquely to a single vertex.

loop

a relationship that connects a node to itself.

neighbor

of node: another node, connected by a common relationship; of relationship: another relationship,connected to a common node.

node*

data record within a data graph; contains an arbitrary collection of properties. Nodes may havezero, one, or more labels and are optionally connected by relationships. Similar to vertex.

null

NULL is a special marker, used to indicate that a data item does not exist in the graph or that thevalue of an expression is unknown or inapplicable.

operator

there are three categories of operators in Cypher:

1. Arithmetic, such as +, /, % etc.;

2. Logical, such as OR, AND, NOT etc.; and

3. Comparison, such as <, >, = etc.

outgoing relationship

pertaining to a directed relationship: from the point of view of a node n, this is any relationship rleaving n, exemplified by (n)-[:r]→(). The opposite of incoming relationship.

pattern graph

graph used to express the shape (that is, connectivity pattern) of the data being searched for in thedata graph. This is what MATCH and WHERE describe in a Cypher query.

path*

collection of alternating nodes and relationships that corresponds to a walk in the data graph.

parameter

named value provided when running a statement. Parameters allow Cypher to efficiently re-useexecution plans without having to parse and recompile every statement when only a literal valuechanges.

predicate

257

expression that returns TRUE, FALSE or NULL. When used in WHERE, NULL is treated as FALSE.

projection

an operation taking result rows as both input and output data. This may be a subset of thevariables provided in the input, a calculation based on variables in the input, or both. The relevantclauses are WITH and RETURN.

property*

named value stored in a node or relationship. Synonym for attribute.

property graph

a graph having directed, typed relationships. Each node or relationship may have zero or moreassociated properties.

query*

statement that reads or writes data from the database

relationship*

data record in a property graph that associates an ordered pair of nodes. Similar to arc and edge.

relationship type

marks a relationship as a member of a named subset. A relationship must be assigned one andonly one type. For example, in the Cypher pattern (start)-[:TYPE]→(to), TYPE is the relationshiptype.

result row

each query returns an iterator of result rows, which represents the result of executing the query.Each result row is a set of key-value pairs (a record).

rollback

abort of the containing transaction, effectively undoing any changes defined inside the transaction.

schema

persistent database state that describes available indexes and enabled constraints for the datagraph.

schema command

statement that updates the schema.

statement

text string containing a Cypher query or command.

type

types classify values. Each value in Cypher has a concrete type. Supported types are:

• string,

• boolean,

• the number types (double, integer, long),

• the map types (plain maps, nodes, and relationships),

• and collections of any concrete type.

The type hierarchy supports several other types (for example, any, scalar, derived map, collection).These are used to classify values and collections of values having different concrete types.

258

transaction

A transaction comprises a unit of work performed against a database. It is treated in a coherentand reliable way, independent of other transactions. A transaction, by definition, must be atomic,consistent, isolated, and durable.

transitive closure

of a graph: is a graph which contains a relationship from node x to node y whenever there is adirected path from x to y; For example, if there is a relationship from a to b, and another from b toc, then the transitive closure includes a relationship from a to c.

undirected relationship

a relationship that doesn’t have a direction. The opposite of directed relationship.

vertex

graph theory: the fundamental unit used to form a mathematical graph (plural: vertices). See node.

259

Chapter 28. LicenseCreative Commons 3.0

You are free to

Share

copy and redistribute the material in any medium or format

Adapt

remix, transform, and build upon the material

for any purpose, even commercially.

The licensor cannot revoke these freedoms as long as you follow the license terms.

Under the following terms

Attribution

You must give appropriate credit, provide a link to the license, and indicate if changes were made.You may do so in any reasonable manner, but not in any way that suggests the licensor endorsesyou or your use.

ShareAlike

If you remix, transform, or build upon the material, you must distribute your contributions underthe same license as the original.

No additional restrictions

You may not apply legal terms or technological measures that legally restrict others from doinganything the license permits.

Notices

You do not have to comply with the license for elements of the material in the public domain or whereyour use is permitted by an applicable exception or limitation.

No warranties are given. The license may not give you all of the permissions necessary for yourintended use. For example, other rights such as publicity, privacy, or moral rights may limit how youuse the material.

See http://creativecommons.org/licenses/by-sa/3.0/ for further details. The full license text is availableat http://creativecommons.org/licenses/by-sa/3.0/legalcode.

260

http://creativecommons.org/licenses/by-sa/3.0/

http://creativecommons.org/licenses/by-sa/3.0/legalcode

The Neo4j Developer Manual v3 - we-yun.com 3.0 Docs/neo4j-developer... · License: Creative Commons...

Documents

Transcript of The Neo4j Developer Manual v3 - we-yun.com 3.0 Docs/neo4j-developer... · License: Creative Commons...